Upstream contribution evidence — @tonydzi

Snapshot: 2026-10-07 · source: public GitHub API · regenerated weekly.

This page exists so the contribution record stays checkable independently of any one profile page. Everything below is a link into someone else's repository: their issue tracker, their pull requests, their release notes.

Counts are split on purpose. Opening a pull request in someone else's project and commenting in someone else's thread are different things, and adding them together would overstate the first.

Pull requests opened in other people's repos	148
…of which merged	52
Issues opened in other people's repos	64
Other people's threads participated in	410
Reviews on other people's PRs	74
Distinct repositories	244
Replies received from maintainers and participants	1996
Release notes crediting this account	21
Files in other repos crediting this account	66

Credited in other projects' release notes

cristicretu/diri — v0.5.1 · 2026-08-14 · release notes

• Add a keyboard shortcuts reference (#52) — @tonydzi
• Sample the stack when the skipped tests hang on a runner (#26) — @tonydzi @alexapvl, @bqx619, @cristicretu, @lucasAguiar11, @tonydzi — and @imgbot.

mixelpixx/Konnect — v0.4.0 · 2026-08-15 · release notes

• The plugin no longer deletes another KiCAD session's server record, and reaps the server it started when KiCAD exits (#199, @tonydzi — partial coverage of #103; the IPC entrypoint still needs its own fix).

• fix(plugin): keep the server PID record correct across sessions and reap on exit by @tonydzi in mixelpixx/Konnect#199
• @tonydzi made their first contribution in mixelpixx/Konnect#199

agno-agi/agno — v3.0.0a3 · 2026-08-21 · release notes

• fix: repair four imports that do not resolve in cookbooks by @tonydzi in agno-agi/agno#9498
• @tonydzi made their first contribution in agno-agi/agno#9498

qualixar/superlocalmemory — v4.1.0 · 2026-08-23 · release notes

Reported by @tonydzi (#122).

basicmachines-co/basic-memory — v0.23.0 · 2026-08-24 · release notes

• fix(integrations): stop bm subprocesses inheriting Hermes's Python env by @tonydzi in basicmachines-co/basic-memory#1179
• @tonydzi made their first contribution in basicmachines-co/basic-memory#1179

agno-agi/agno — v3.0.0 · 2026-08-24 · release notes

• fix: repair four imports that do not resolve in cookbooks by @tonydzi in agno-agi/agno#9498
• @tonydzi made their first contribution in agno-agi/agno#9498

michellzappa/headroom — v2.0.8 · 2026-08-26 · release notes

(#28, reported by @tonydzi). Claude Code writes one JSONL line per content

Lyellr88/marm-memory — v2.45.0 · 2026-08-29 · release notes

The README published to PyPI carried 17 links written relative to the repository root, a position neither surface that renders that file ever occupies. Reported and fixed by @tonydzi.

• docs(pypi): make the PyPI README's links absolute so they resolve by @tonydzi in Lyellr88/marm-memory#180
• @tonydzi made their first contribution in Lyellr88/marm-memory#180

QwenLM/qwen-code — cua-driver-rs-v0.20.3 · 2026-09-01 · release notes

• fix(core): do not claim a question host in stream-json direct mode by @tonydzi in QwenLM/qwen-code#9414
• @tonydzi made their first contribution in QwenLM/qwen-code#9414

Lyellr88/marm-memory — v2.46.2 · 2026-09-02 · release notes

The 8 tests in tests/test_docker_transports.py that actually start a container previously only ran in the release workflow, which triggers on a version tag, so a Docker runtime regression was caught after a release was already tagged rather than on the PR that caused it. Contributed by @tonydzi. scripts/test-scripts/smoke_embedding_chunking.py had been broken since chunking was split into separate memory and document profiles: it imported constants that no longer existed and called _chunk_text without the keyword arguments it now requires. Contributed by @tonydzi.

• fix(scripts): repair smoke_embedding_chunking against the v2.14.0 chunk profiles by @tonydzi in Lyellr88/marm-memory#181
• ci: build the image on PRs and run the docker-marked tests (#170 item 1) by @tonydzi in Lyellr88/marm-memory#183

modelcontextprotocol/go-sdk — v1.8.0-pre.1 · 2026-09-04 · release notes

• docs: make the protocol.md table of contents resolve by @tonydzi (#1196)
• docs: fix the streamable client snippet to use (*Client).Connect by @tonydzi (#1142)
• docs: quick start go get leaves the module unbuildable by @tonydzi (#1148)
• docs: fix four dead links to the Multi Round-Trip Requests section by @tonydzi (#1159)

• @tonydzi made their first contribution in modelcontextprotocol/go-sdk#1142

modelcontextprotocol/go-sdk — v1.8.0-pre.2 · 2026-09-04 · release notes

• docs: make the protocol.md table of contents resolve by @tonydzi (#1196)
• docs: fix the streamable client snippet to use (*Client).Connect by @tonydzi (#1142)
• docs: quick start go get leaves the module unbuildable by @tonydzi (#1148)
• docs: fix four dead links to the Multi Round-Trip Requests section by @tonydzi (#1159)

• @tonydzi made their first contribution in modelcontextprotocol/go-sdk#1142

mixelpixx/Konnect — v0.11.1 · 2026-09-07 · release notes

• fix(placement): stop scoring connector filter caps as decoupling defects by @tonydzi in mixelpixx/Konnect#416

modelcontextprotocol/go-sdk — v1.8.0 · 2026-09-14 · release notes

• docs: make the protocol.md table of contents resolve by @tonydzi (#1196)
• docs: fix the streamable client snippet to use (*Client).Connect by @tonydzi (#1142)
• docs: quick start go get leaves the module unbuildable by @tonydzi (#1148)
• docs: fix four dead links to the Multi Round-Trip Requests section by @tonydzi (#1159)

• @tonydzi made their first contribution in modelcontextprotocol/go-sdk#1142

Lyellr88/marm-memory — v2.48.1 · 2026-09-16 · release notes

• test(docker): the healthcheck could never report unhealthy (#170 item 5) by @tonydzi in Lyellr88/marm-memory#185

mixelpixx/Konnect — v0.12.0 · 2026-09-17 · release notes

• #505 — Discover KiCad sockets from metadata without probing by connection. — @tonydzi
• #442 — Record running servers and sweep dead process records at startup. — @tonydzi

Lyellr88/marm-memory — v2.49.0 · 2026-09-18 · release notes

• test(docker): exposed mode authenticates, and MARM's managed Docker path never reaches the keyless fallback (#170 item 5) by @tonydzi in Lyellr88/marm-memory#192

chigwell/telegram-mcp — v3.2.42 · 2026-09-19 · release notes

• feat: make per-tool extension allowlists configurable via TELEGRAM_FILE_EXTENSIONS (closes #226) by @tonydzi in chigwell/telegram-mcp#227
• @tonydzi made their first contribution in chigwell/telegram-mcp#227

langroid/langroid — 0.68.1 · 2026-09-23 · release notes

are being walked through opening a PR (#1148, by @tonydzi).

microsoft/semantic-kernel — python-1.45.0 · 2026-10-06 · release notes

• Python: pin the validated address for OpenAPI plugin requests by @tonydzi in microsoft/semantic-kernel#14371
• @tonydzi made their first contribution in microsoft/semantic-kernel#14371

langroid/langroid — 0.68.7 · 2026-10-07 · release notes

Task.run(turns=...) or Settings.max_turns. Thanks to @tonydzi, whose

Credited in other projects' files

Lines maintainers wrote into their own changelogs and contributor lists.

eugeniughelbur/obsidian-second-brain · CHANGELOG.md

• The write-time hook reported payload keys: none for a payload that had a path key one level up (#171, reported by @tonydzi). A payload like {"tool_name":"Write","file_path":"..."} still exited 1 and stayed fail-closed, but the diagnostic listed only keys under tool_input and args. It n …
• validate-ai-first.sh exited 0, silently, when a write payload named a tool but carried no path key it knew (#171, the open item from @tonydzi's codex-cli reports). The hook read file_path and filePath (and args.*); a host that sends the path under another key (path, uri) got exit 0 …

basicmachines-co/basic-memory · CHANGELOG.md

@tonydzi for the report and fix (#1617, landed as #1619).

Lyellr88/marm-memory · CONTRIBUTORS.md

• Anton Dziatkovskii (@tonydzi) —

Lyellr88/marm-memory · CHANGELOG.md

The 8 tests in tests/test_docker_transports.py that actually start a container previously only ran in the release workflow, which triggers on a version tag, so a Docker runtime regression was caught after a release was already tagged rather than on the PR that caused it. Contributed by [@tonydzi]( … scripts/test-scripts/smoke_embedding_chunking.py had been broken since chunking was split into separate memory and document profiles: it imported constants that no longer existed and called _chunk_text without the keyword arguments it now requires. Contributed by [@tonydzi](https://github.com/to … The README published to PyPI carried 17 links written relative to the repository root, a position neither surface that renders that file ever occupies. Reported and fixed by @tonydzi.

qualixar/superlocalmemory · CHANGELOG.md

Reported by @tonydzi (#122).

michellzappa/headroom · CHANGELOG.md

(#28, reported by @tonydzi). Claude Code writes one JSONL line per content

majiayu000/claude-skill-registry-data · other/llll/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-ml/rr/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/tt-probe/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-ml/screenpipe-worklog/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/screenpipe-recall/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/screenpipe-meeting-prep/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/find/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/agenda/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-ml/screenpipe-shareable-recap/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-ml/chat/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-ml/screenpipe-research-synthesis/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · productivity/screenpipe-focus-review/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-llm/secondop/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/alpha-review/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/declined/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/five-hard/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/tg-post/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-ml/resume-last/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-ml/dedup/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/precedent/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-ml/chat-search/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/wisdom-distill/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · productivity/screenpipe-durable-learning/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/skill-gap/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/hubrun/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/portret/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/screenpipe-meeting-follow-up/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · bash/reboot/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-llm/bible/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · bash/worth/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-ml/goblin/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/wow/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/mine-channel/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/community-alpha/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/cofounder-watch/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · bash/fb-post/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/bold-followup/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/hk/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/comments/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · development/fleet/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/guide/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/cofounder/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/tg-check/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · ai-ml/sync-sessions/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · bash/03/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/quarantine/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/promises/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · productivity/intro/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/episode/SKILL.md

• canonical = the GitHub markdown on every platform (a single source of truth for AI). Repo: github.com/tonydzi/clawrush. Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/speak-as/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/telegram-mcp-kit/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X [аккаунт]

majiayu000/claude-skill-registry-data · other/chip/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/deflate/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/ask-any/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/fireflies-sync/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/bus/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/tg-slot/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/gitbook-import/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · other/health-sync/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

majiayu000/claude-skill-registry-data · context-management/search/SKILL.md

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything i … Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tony … Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

What maintainers replied

Replies in threads we opened, plus replies elsewhere that name this account. Quotes are trimmed; every one links to the original.

anthropics/claude-code issue #91188 — Feature request: make the auto-memory MEMORY.md compaction reminder threshold configurable

Role here: our comment in their thread.

DanceNitra, 2026-09-01:

@niels-roest, your 17.1KB and the ~70% you read off it are both exact. The constants are literals, and there is a second half that changes what your knob would have to do. Where the numbers come from. The reminder builds one entry per dimension, each carrying a fraction of its own cap, and reports whichever is proportionally closest: ```js var ljo=0.8,HUn=0.7; function xut(e){ let n=[{frac:e.sizeBytes/e.byteCap,dimension:"bytes", capDesc:Ut(e.byteCap),targetDesc:Ut(Math.floor(e.byteCapHUn))}]; if(e.lineCap!==void 0&&e.lineCount!==void 0) n.push({frac:e.lineCount/e.lineCap,dimension:"lines", capDesc:${e.lineCap}-line,targetDesc:`${Math.floor(e.lineCapHUn)} lines`}); let{frac:r,...o}=n …

DanceNitra, 2026-09-03:

@niels-roest, the reminder can be made to fire later. The branch that does it is one none of us had read, and finding it means retracting a sentence of my own. On 1 September I wrote that "raising the configured value raises the number you are told to aim for". That is wrong on the default path. The four fields at the call site go into one function, and that function is a selector: js function IHe({rawSizeBytes:e,surfaceCap:n,splicedSizeBytes:r,spliceCap:o,spliceActive:d}){ return n!==void 0 && (!d || e/n >= r/o) ? {sizeBytes:e,byteCap:n} : {sizeBytes:r,byteCap:o} } !d short-circuits, so the two branches behave differently. With spliceActive true, the configured cap wins only whil …

gastownhall/beads issue #5877 — Proposal: Memory Beads

Role here: our comment in their thread.

The1nk, 2026-09-22:

@tonydzi — you found a real hole in our strongest number. The honest response is to go and measure it rather than defend it, so I did, and you were right in a way that is slightly worse than you could have seen from outside. ## Our override count was scoped to one surface, and we never said so bd-notes-clobber-guard is wired PreToolUse on Bash only, and it reads exactly one field — tool_input.command. That string is everything it knows about the world. So rather than reason about bypasses I fed them to it, with the real refusals as a positive control and the log redirected so the probe could not contaminate the number it was testing: ``` deny gc bd update bl-x --notes "wipes the th …

The1nk, 2026-09-21:

@tonydzi Your one line — a rule does not count as landed until you can name the thing that calls it — is the control we converged on too. I can put a number on the failure it prevents, because we measured the always-loaded plane failing while the rule was sitting in it. It is @seanmartinsmith's third canonical rule, --notes replaces the whole field. The ban was written in three files here. Measured across one store on 2026-08-20, with bd's own warning already firing: 309 note-writes, 33 of them destroyed another actor's note, and 23 of those 33 were a mayor's scoping or ruling wiped by the worker that then picked the bead up. Single losses of 13.7 KB, 11.8 KB, 9.7 KB — the same aet …

bytedance/deer-flow PR #5843 — fix(docs): harness docs document an async client API that does not exist

Role here: our pull request.

WillemJiang, 2026-09-25:

@tonydzi Thanks for your contribution, please click the CLA assistant button to sign the CLA then we can merge your code.

WillemJiang, 2026-09-27:

@tonydzi, please click the CLA button posted by the CLA assistant in the PR comments to sign the CLA.

run-llama/llama_index issue #23122 — Integration proposal: memory store with verified transfer — export/import sealed memory bundles

Role here: our comment in their thread.

Kaushalt2004, 2026-10-03:

Both of these are right, and they're different sizes of right. On the chain test — yes, and I'd rather you write it than me. You spotted it, you built the vector it protects, and the shape you describe is exactly the one this thread has earned: enumerate the links by name, assert each exists at its registered path, runs, and fails when its own subject is disabled. The third clause is the one that makes it a test rather than an import check, and it's the clause most people skip. tests/ next to seam_mutation.py, failing loudly on structural drift with the name of the link it lost — that's the right home and the right failure mode. Land it and I'll verify it the same way as the last t …

eddyflores100-lang, 2026-09-30:

@tonydzi — took your suggestion to its conclusion, because the second-order edge you named was the real indictment: a red whose message accuses the source of drifting when the event was "someone defeated this first" is a tripwire that can't be trusted to describe its own firings. The tripwire is now semantic, like its sibling: cognicore-dev/cognicore-env#145. The mutation is a runtime patch at the recall seam — every query "finds" every stored entry, so dark is always empty and the verdict can never fire, while if dark: stays byte-identical. Your mutation B, expressed as a patch instead of a source edit. What falls out of that: - …

google/adk-go PR #1299 — fix(configurable): apply config_path containment to workflow node refs

Role here: our pull request.

karolpiotrowicz, 2026-09-01:

The containment fix itself holds up. I checked out the branch and drove the traversal through FromConfig rather than the helper: on the merge-base a workflow edge naming ../../outside.yaml loads and builds the workflow, and on this branch both that and the absolute-path form are refused. The full suite shows no new failures against the merge-base, and the routing through one shared helper ahead of both the read and the cache lookup is the right shape. One thing needs settling before this merges, and it is the behaviour change @jjsasha63 already flagged as needing a wider note — it is wider than it currently reads. ## Refusing every link, not just escaping ones, breaks whole config layout …

karolpiotrowicz, 2026-09-04:

Every ask from my last review is answered, and I checked each one against the branch rather than reading the response commits. The trade is stated in the description now, with the ConfigMap consequence and two concrete routes for an operator on such a mount. TestResolveConfigReferenceRefusesLinksThatStayInside pins the behaviour the change deliberately chooses, and it discriminates: reverting the walk to resolve-the-link-and-refuse-only-if-it-escapes fails that test and no other, with the escaping-symlink case still green. The EvalSymlinks block now has both a comment that describes what it actually does and a test that fails if you delete it. ENOTDIR, the empty-reference diagnostic, the …

microsoft/agent-framework PR #7581 — Python: bind tool-approval responses to surfaced approval requests (#7383)

Role here: our pull request.

eavanvalkenburg, 2026-08-26:

@tonydzi thanks for working on this, i think the overall idea makes sense, the thing that is tricky here is that there is only so much we can do, because if a attacker has access to your session store they can change that version of the FCC and have it executed, but at some point we need to trust something, and this does add a additional layer that is less user controlled (session vs input message), so a improvement nonetheless. So let's get this in shape and then I will do a deeper dive. I'm also thinking about ID's, whether call_id is the right one, since some chat clients do not generate them...

eavanvalkenburg, 2026-09-01:

@tonydzi, thank you for putting this draft together and especially for reproducing and documenting the failure modes raised in review. The core record/rebind/consume mechanism landed independently in #7631, so this branch is now superseded. I'm closing it in favor of #7988, which carries forward the remaining work around stable local approval occurrence identities, replay coverage, provider correlation, and the trusted-storage boundary. The investigation and review findings here directly shaped that PR—thank you.

tenequm/pond PR #321 — fix(schedule): report a systemd timer with no next elapse as broken

Role here: our pull request.

tenequm, 2026-09-30:

[Blocking] The second remedy here, pond schedule start, is a no-op in every state this detector fires on. The detector only runs after systemctl --user is-enabled succeeded, and in the field case nothing edited the unit files - which is exactly start_systemd's early-return condition (schedule.rs:822, if unchanged && systemd_timer_enabled()). The user sees already scheduled (every 5m) and the timer stays dead. Even past that check, enable --now issues a start job on an already-active (elapsed) unit - a no-op: systemd's timer_start asserts TIMER_DEAD/TIMER_FAILED, and daemon-reload coldplug re-enters ELAPSED verbatim (checked against timer.c). And for a stopped or fai …

tenequm, 2026-09-30:

Sharp find, solid evidence - the red-first tests and the real-host capture that overturned the issue's suggested NextElapseUSecRealtime reading are exactly how this kind of fix should arrive. Requesting changes on one defect: the message's pond schedule start remedy is a no-op in every state this detector fires on (details inline), which collides with pond's rule that a detected bad state names the exact working recovery command. 3 blocking asks (one is comment-only) and 3 non-blocking suggestions, all inline. cargo fmt --check / cargo clippy --all-targets -- -D warnings / cargo test are green locally on this branch (macOS; CI had not yet been approved to run when I reviewed). Pond …

microsoft/semantic-kernel PR #14199 — Python: Add experimental FunctionAuthorizationFilter for auto function invocation (runtime authorization, argument-bound approvals)

Role here: our pull request.

babyblueviper1, 2026-07-27:

Reviewed the actual implementation, not just the description — the args_digest binding is real and sound: sha256(function_name | args_digest | principal | policy_digest) over a structurally type-tagged canonicalization (the _canonicalize method), which is the right defense against a hostile __str__ producing a digest collision. That's a genuinely non-obvious detail to get right and it's handled correctly here. This directly answers the question I raised on #14072 about whether a dispatcher verifies the stored draft against what actually executes — it does, cryptographically, by construction: transfer(amount=10) approved can never authorize transfer(amount=10000) executed, because …

babyblueviper1, 2026-07-27:

Worth pulling the automated DevFlow finding above into the same class of gap this thread's been naming, since it's a real one and distinct from the attribution question: the args_digest binding I checked earlier is sound (replay/tamper is structurally prevented), but that's orthogonal to when enforcement actually happens. asyncio.gather(...) dispatches every tool call in a model response concurrently and only inspects terminate after the whole batch resolves — so a pending_approval verdict on call N doesn't stop sibling calls in the same batch from having already executed by the time terminate is checked. The digest math being unforgeable doesn't help if the enforcement point and …

pydantic/pydantic-ai issue #8551 — An uninstrumented delegate's usage is credited to the caller's agent run span

Role here: our issue report.

yetuge, 2026-09-20:

Picking up @84dnnvbdvp-debug's durable edge and this thread's open caveat: I built a local harness for it, and the replay leak reproduces on main (c4898ab). That discharges the "durable-execution interaction unverified" caveat @tonydzi left open — the run-entry ownership move now has a concrete success criterion on this path. Shape: instrumented parent (FunctionModel) → uninstrumented durable delegate, where the delegate carries a durable capability operation that folds a usage delta across the boundary. Two consecutive parent.run() calls, so the second one replays the delegate's operation instead of re-executing its body. Span attributes read off an in-memory OTel exporter: ``` run1 (fi …

qaisermehdi3-coder, 2026-09-23:

@tonydzi thanks for testing it again instead of agreeing. two small things from my side. 1. the marker summing into RunUsage is the part that matters most. i hit this in my own tool tonight: calls were marked unknown, but the per task average still used them. two workflows doing identical work, one with a call that returned no usage, showed $1.00 per task vs $2.00. the half measured one looked twice as cheap. marking the request is not enough, the total it rolls into has to carry the mark, and yours does. 2. on the streaming path: in openai chat streams every chunk except the last has usage None, and without include_usage all of them do. you said incr only counts requests because _map_usage …

Jason-Vaughan/TangleClaw issue #1400 — [bug] Primary-checkout guard hook fails silently due to || true

Role here: our comment in their thread.

Jason-Vaughan, 2026-09-23:

Thanks @tonydzi — this was genuinely useful, and it prompted me to audit that path more closely. I am removing the outer || true, but with one important distinction: the primary-checkout guard is a safety interlock rather than a security boundary, so an internal guard failure still has to remain non-blocking. A deliberate refusal and a guard crash are separate outcomes; I don't want a bug in the interlock itself to strand the Claude Code session it is protecting. The false positive case you described turns out not to apply to TangleClaw's implementation. File and notebook operations don't infer targets from command text. They arrive as structured paths, are resolved to absolute/symlink-res …

Jason-Vaughan, 2026-10-01:

@tonydzi This is exactly the implementation scar tissue that keeps a good acceptance criterion from becoming a decorative one. All three points are accepted. I tightened the issue body so the requirements now say explicitly: - liveness is measured against an independently produced activity signal that the guarded hook cannot influence, not wall-clock age; - an idle install stays quiet only when both expected activity and heartbeat movement stay quiet; - permanent silence exemptions are prohibited, and any unavoidable exclusion needs a bounded reason plus an expiry that becomes actionable; - --self-test must traverse the real installed hook entrypoint and prove the outer heartbeat advanced, …

Lyellr88/marm-memory PR #192 — test(docker): exposed mode authenticates, and MARM's managed Docker path never reaches the keyless fallback (#170 item 5)

Role here: our pull request.

Lyellr88, 2026-09-17:

@tonydzi Thank you for the continued work on this. The revised tests clearly separate the managed keyless-start security contract from persistence, cover the exposed-network path without relying on host behavior that is not portable, and protect the probes from ambient proxy interference. Everything is green and the review is complete. Merging this now. Please keep an eye on the open issues if another area interests you.

Lyellr88, 2026-09-16:

I apologize for the delay; I've been working on a new build plus the power outage I dealt with pulled my focus in multiple directions. Thanks for the focused Docker coverage and for including both CI and mutation-test evidence. I verified the tests run in the Docker CI job and agree that the intended auth cases are worth covering. 1. marm-mcp-server/tests/test_docker_transports.py:659-711: the no-key test currently requires a persisted /home/marm/.marm/.env key via docker exec, but the security property under review is that the loopback fallback is unreachable. marm_mcp_server/config/api_key_bootstrap.py:53-86 can retain a generated key in memory while declining persistence if secure …

QwenLM/qwen-code PR #13312 — fix(mcp): name the capped server timeout in App read timeout warnings

Role here: our pull request.

wenshao, 2026-10-06:

Maintainer verification (round 2): findings — 47/47 scripted assertions passed (wire A/B 32, TUI A/B 2, unit gates 2, mutation matrix 4, static gates 4, merge check 1, censuses 2). The pass count confirms every predicted behavior — including the prediction that R1-1 is real. Verified head 5245ba6fb6913d6c56c667fc36c8a7d5353d48ed against merge-base eb0b79c5b87d781cddc65c28e70955404eb928d3; trial merge into current main (fd4af70) is conflict-free and the merged tree passes the affected suite 366/366.

中文摘要

第二轮验证结论：有发现（findings）——R1-1 在真实线级环境被端到端确认，合并决定取决于默认配置臂的措辞取舍（作者已明确把该决定留给维护者）。 - 历史发现状态：R1-1 仍然成立（head 自上一轮以来未变，本轮全部重测）；R1-4 部分需要更正——其精确变异 …

qwen-code-review-bot, 2026-10-07:

[Suggestion] R1-4: A written-but-non-finite server timeout becomes detectable at this line, but it is discarded with no diagnostic at any level — and the arm this diff adds for that case tells the operator appResourceTimeoutMs (default 10000 ms), where the word "default" asserts they wrote nothing. defaultTimeoutMs is computed by a bare boundedAppLimit call (:882) that bypasses the appResourceLimit wrapper (:828-844), whose docblock says the diagnostic exists precisely because such a value "would be dropped without a trace while the limit warning names the key". The two sibling App keys do get an Ignoring non-finite MCP App resource limit … line, pinned by `expectDiscarde …

anthropics/anthropic-sdk-python PR #1820 — Fix streaming accumulator crash when message_start omits usage

Role here: our comment in their thread.

PiedPiper911, 2026-08-27:

@tonydzi — thank you, again, for the re-measurement. All three findings were correct, and I have acted on all of them. Pushed as commit : 1. None coercion (the real fix). Both accumulators ( + ) now coerce a missing to before /. Your reproduction was exactly right: bypasses validation, so on #1806's reported shape the field lands as on a required — an AttributeError traded for a TypeError one frame further from the cause. 2. Fixture restored to the reported shape. now ends with — no , matching the actual repro in #1806. 3. Discriminating assertion + non-strict client. The tests now assert (which fails on the old implementation) and build a non-strict locally — you were right that …

PiedPiper911, 2026-08-24:

Thanks for the drive-by review, tonydzi — appreciate the triage and the honest read on both PRs. I've since synced this branch with the latest main (merge commit 1316de75 + a line-ending cleanup), so the diff is now just the actual fix: 4 files instead of the 60-file noise from the stale base. The fix itself is unchanged — when message_start omits usage, the accumulator constructs it from the first message_delta instead of crashing (#1806), with sync + async regression tests (new missing_usage_response.txt fixture). On the overlap with #1815: you're right that @chenlichao opened it ~13h earlier, and your point about the ideal merge (this fix + a solid test setup) is fair. Both PRs …

anthropics/claude-code issue #90542 — [BUG] A complete CLAUDE.md rule contract governed nothing: 9 fabricated causes, stale state asserted as current, acceptance step silently skipped across a 4.5h session

Role here: our comment in their thread.

rulereceipt, 2026-09-14:

@stonianua — measured, and you were right about the scope. Branch policy took the first backtick literal in any rule containing the word "branch" and called it a branch name. Across 559 rules files, 37 of the 70 rules routed there (52.9%) were handed something that cannot be a branch — a naming template squad/{issue-number}-{kebab-case-slug}, whole commands, one entire fenced markdown block. The report said: no git command targeted the git push --force branch this session. Now 0 of 45. Claim-evidence was the same fault. Reporting verb, evidence noun, done-word, each anywhere in the rule, independently. A 944-character rule about writing status updates satisfied all three across three par …

rulereceipt, 2026-09-16:

@stonianua — named and fixed, and the harness had a flaw of its own. That one first, because it changes my last number. I said it pins its inputs. It picks the largest sessions deterministically, which is not the same as reading the same bytes twice — and the largest here include the session doing the measuring, appended to while it runs. 26 texts became 27 with no code change, because 2MB of transcript arrived in between. Each input now prints the sha256 of what was read. Pinned: 30 of 2,795 reports, 1.1%, 22 texts. The causes. The heredoc guard existed and was wired into one reader — I wrote it for test-command detection and never applied it to file mutations, so a command editing landing/ …

google-gemini/cookbook PR #1296 — Add example: check citation faithfulness in RAG

Role here: our pull request.

kkorpal, 2026-08-06:

Hi @Palo-Alto-AI-Research-Lab While testing the latest changes, I ran into a couple of structural errors during execution: A TypeError is triggered during the client.interactions.create() execution step. A ValidationError follows during the subsequent Pydantic model_validate_json() parsing loop. Please review the official Google AI Studio Interactions API Overview and the Interactions API Reference Guide to confirm how parameter structures, schema types, and response block formats must be formatted for this endpoint

kkorpal, 2026-07-27:

Hi @Palo-Alto-AI-Research-Lab , Before we move forward, could you take a quick look at the failing checks on your PR? It looks like the Google CLA needs to be signed, and the notebook format and lint checks are currently failing. Definitely check out the automated bot suggestions on the PR thread. Once those are all green, we can get this moving. Thanks

Full index

Every position we opened, newest first. Full bodies and every reply are in evidence-index.json.

Repository	#	Kind	State	Title
mariagorskikh/open-instinct	#12	PR	merged	fix(policy): blocked merchants are missed unless the name is a top-level key
mariagorskikh/open-instinct	#11	PR	merged	ci: build before typecheck so the check job can pass
QwenLM/qwen-code	#13616	PR	open	test(desktop): pin both Linux release legs to the ubuntu-22.04 glibc floor
DasterProkio/awesome-ai-companion	#43	PR	closed	Add The Bible Framework (persona as versioned rules) to Memory & Identity
OWASP/www-project-agentic-skills-top-10	#89	PR	open	proposals/ast-fixture-corpus: AST09 actor-tamper vector (requested in #81) + two JCS canonicalization vectors
langroid/langroid	#1191	PR	closed	docs: task-termination guide passes single_round/done_if_response to TaskConfig, where they are silently ignored
ctxrs/ctx	#1117	PR	open	docs: fix broken link to the v0.4.0 Sift benchmark in README
pydantic/pydantic-ai	#9835	PR	closed	fix(bedrock): keep every nova_code_interpreter_result toolResult content block
QwenLM/qwen-code	#13312	PR	open	fix(mcp): name the capped server timeout in App read timeout warnings
screenpipe/screenpipe	#7401	issue	open	Windows: communications-output loopback (USB speakerphone) dies 200ms after start with 'disconnected', while /health keeps reporting the device active — call far-end silently lost
pydantic/pydantic-ai	#9444	issue	closed	A provider response with no usage object is recorded as zero tokens, not as unknown
tenequm/pond	#321	PR	merged	fix(schedule): report a systemd timer with no next elapse as broken
basicmachines-co/basic-memory	#1617	PR	closed	fix(mcp): report the real checksum from write_note and edit_note (#1586)
ChromeDevTools/chrome-devtools-mcp	#2852	PR	open	docs: correct CLI availability of extension tools
headroomlabs-ai/headroom	#3832	PR	merged	docs: mark HeadroomClient provider as required, not auto-detected
QwenLM/qwen-code	#12875	PR	open	fix(core): add opt-in failMode: "closed" for PreToolUse command hooks
bytedance/deer-flow	#5843	PR	merged	fix(docs): harness docs document an async client API that does not exist
deepset-ai/haystack	#12889	PR	closed	fix: merge generation_kwargs tools with the component's own tools in OpenAIChatGenerator
anthropics/claude-code	#95920	issue	open	Scheduled tasks: nextFire returns a fabricated "now + 365 days" when no match falls inside its 366-day scan window
langroid/langroid	#1148	PR	merged	docs: fix dead github-cli link in CONTRIBUTING
browser-use/browser-use	#5867	PR	open	docs(actor): fix drag_to parameter name in Element API reference
teng-lin/notebooklm-py	#2432	issue	closed	wait_for_completion returns REMOVED for generations that completed — absence of task_id from the studio list is inferred, never verified (0.8.2)
QwenLM/qwen-code	#12422	PR	open	fix(acp): report max_tokens instead of end_turn on unresolved output truncation
LambdaLabsML/the_lab.api	#7	issue	open	optimization/test_project: baseline fails its own accuracy gate on 3 of 7 kernels
LambdaLabsML/the_lab.api	#6	PR	open	fix(dashboard): install node dependencies before building
LambdaLabsML/the_lab.api	#5	PR	open	fix(mcp): resolve the API port from runtime.json instead of assuming 8000
BerriAI/litellm	#42172	issue	open	[Bug]: anthropic/ with a third-party api_base receives the client's Claude subscription OAuth token instead of the deployment's configured api_key (with forward_llm_provider_auth_headers either on or off)
pydantic/pydantic-ai	#8551	issue	closed	An uninstrumented delegate's usage is credited to the caller's agent run span
chigwell/telegram-mcp	#227	PR	merged	feat: make per-tool extension allowlists configurable via TELEGRAM_FILE_EXTENSIONS (closes #226)
mem0ai/mem0	#7361	PR	open	docs(llms): Sarvam sampling params go inside the model dict
pydantic/pydantic-ai	#8451	issue	closed	Concurrent delegate runs still over-report usage on their agent-run spans after #8417
basicmachines-co/basic-memory	#1578	issue	open	Harden HTTP/SSE transport: basic-memory mcp defaults to 0.0.0.0 with no auth (contradicts SECURITY.md)
oraios/serena	#2048	PR	closed	docs(clients): fix CC system prompt override command (prompts print-cc-system-prompt-override)
DeusData/codebase-memory-mcp	#2225	PR	merged	docs: MCP tool count is 17; document get_file_outline, compare_graphs, check_index_coverage
LambdaLabsML/the_lab.api	#4	PR	open	fix(mcp): keep the bridge importable under the system python3 (3.9)
cristicretu/diri	#305	PR	merged	engine: sanitize bracketed text in Session::paste_text (#275)
NangoHQ/nango	#7544	PR	closed	feat(integrations): add support for huggingface
NangoHQ/nango	#7543	PR	closed	feat(integrations): add support for cohere
NangoHQ/nango	#7542	PR	closed	feat(integrations): add support for fireworks-ai
NangoHQ/nango	#7541	PR	closed	feat(integrations): add support for together-ai
NangoHQ/nango	#7540	PR	closed	feat(integrations): add support for groq
NangoHQ/nango	#7539	PR	closed	feat(integrations): add support for deepseek
NangoHQ/nango	#7537	PR	closed	feat(integrations): add support for mistral
Kong/kong	#14999	PR	open	fix(ai-proxy): keep empty anthropic tools array
NangoHQ/nango	#7533	PR	open	feat(integrations): add support for lambda-cloud
NangoHQ/nango	#7532	PR	closed	feat(integrations): add support for openrouter
nanbingxyz/5ire	#437	issue	open	Hardening: main BrowserWindow uses webSecurity:false + nodeIntegration:true (Electron security defaults)
openai/openai-agents-python	#5026	issue	closed	Skills frontmatter parser mangles multi-line descriptions (folded >, literal |, wrapped lines) in the skill index
basicmachines-co/basic-memory	#1551	PR	closed	feat(core): accept SQLite URLs in database_url (#539)
microsoft/playwright	#42669	issue	closed	[BiDi] firefox channel="moz-firefox": driver crashes on console log of object with non-primitive toString/valueOf ("Cannot convert object to primitive value")
gklein/fullsend-autopilot	#1	issue	open	unattended merges: how do you reconstruct what the agent did + roll back?
mixelpixx/Konnect	#505	PR	merged	fix(platform): discover KiCad's socket from metadata, never by connecting
elgorro/aiquila	#504	PR	merged	fix(mcp): treat Talk's 304 Not Modified as an empty result
google/adk-go	#1543	PR	open	feat(model/openaimodel): support image/file input via genai InlineData/FileData parts
osaurus-ai/osaurus	#2669	PR	open	docs: fix six relative links that resolve to files that do not exist
areal-project/AReaL	#1685	PR	merged	docs: repoint doc links left behind by the bilingual docs split
zilliztech/memsearch	#725	PR	closed	feat(plugins): make the daily memory filename collision-safe on synced directories
llvm/llvm-project	#221677	PR	open	[Github] Make code-format-helper --extensions ordering deterministic
vansh7nvc/Abstractify	#69	PR	merged	feat(export): add RIS citation exporter for Zotero and Mendeley (#10)
mixelpixx/Konnect	#442	PR	merged	fix(platform): record each server and sweep dead records at startup
mixelpixx/Konnect	#439	PR	closed	fix(drc): name what owns each DRC report item (#413)
mongodb-js/mongodb-mcp-server	#1491	PR	closed	docs: fix Azure deploy paths after the v3 multi-package move
darrenhinde/OpenAgentsControl	#356	PR	open	docs: point plugin setup links at the sections that replaced the removed guides
microsoft/semantic-kernel	#14371	PR	merged	Python: pin the validated address for OpenAPI plugin requests
Lyellr88/marm-memory	#192	PR	merged	test(docker): exposed mode authenticates, and MARM's managed Docker path never reaches the keyless fallback (#170 item 5)
ip2a/mcpstore	#56	issue	open	npm CLI install is broken: README names an unpublished package, and @ip2a/mcpstore ships no bin
bobmatnyc/claude-mpm	#961	PR	open	docs: trusty-memory and trusty-search are cargo crates, not uv tools
mixelpixx/Konnect	#416	PR	merged	fix(placement): stop scoring connector filter caps as decoupling defects
Lyellr88/marm-memory	#185	PR	merged	test(docker): the healthcheck could never report unhealthy (#170 item 5)
google/adk-python	#6980	issue	open	AutoTracingPlugin rebinds what getmembers() returns, not what the class holds: @staticmethod becomes an instance method, @classmethod is never traced, base methods get pinned onto subclasses
microsoft/semantic-kernel	#14355	PR	open	Python: document connector extras and optional dependency compatibility
Lyellr88/marm-memory	#183	PR	merged	ci: build the image on PRs and run the docker-marked tests (#170 item 1)
purarue/google_takeout_parser	#97	issue	open	html activity: file_dt is taken from the file's mtime, which does not survive copying (silent 1h offset)
zackproser/pr-babysitter	#1	issue	open	field notes from a month of running the same class of tool (digest drift, dead-mechanism silence, tombstones)
knowsuchagency/mcp2cli	#107	PR	closed	docs: bake example used an npm package that does not exist
runesleo/x-reader	#25	PR	closed	docs: point Claude Code MCP setup at the file Claude Code reads
topoteretes/cognee-community	#170	PR	closed	feat(connector): Reddit data-source connector (cognee#4811)
anthropics/claude-code	#90962	issue	open	[BUG] Windows: Write tool emits .ps1 without a BOM; PowerShell 5.1 parses it as ANSI and a mojibake quote kills the script - silently when scheduled (0x80070001, no log)
topoteretes/cognee-community	#166	PR	closed	feat(connector): Telegram data-source connector (cognee#4730)
topoteretes/cognee-community	#165	PR	closed	feat(connector): Obsidian vault data-source connector (cognee#4725)
getzep/graphiti	#1810	PR	open	feat(mcp): surface ranker, scores, and invalidated-fact counts in search results (#1645)
Lyellr88/marm-memory	#181	PR	merged	fix(scripts): repair smoke_embedding_chunking against the v2.14.0 chunk profiles
anthropics/claude-code	#90533	issue	open	Scheduled tasks: one string fireAt (written by the app's own update_scheduled_task MCP) makes loadScheduledTasksFromDisk reject the whole registry — all routines silently stop
google/adk-python	#6941	issue	closed	Session id is normalized on write but not on read: after bfeb04c a padded id creates a session that cannot be read, deleted, or re-created (in_memory and sqlite, incl. the adk web / adk run store)
punkpeye/fastmcp	#344	PR	merged	docs(oauth): the "Cannot find module" fix suggested a broken import
Lyellr88/marm-memory	#180	PR	merged	docs(pypi): make the PyPI README's links absolute so they resolve
TsinghuaC3I/Awesome-Memory-for-Agents	#38	PR	merged	Product & Project: add sqlite-graph-memory (wikilink graph + SQLite, no graph DB)
aristoapp/awesome-second-brain	#54	PR	open	watchlist: add CharmOS (humans + agents in one relationship record on a local vault)
Yigtwxx/awesome-claude-multi-agent	#3	PR	closed	Memory, State, and Communication: add claw-consensus (cross-machine consensus + dual-rail bus)
mex-memory/mex	#148	PR	open	fix(cli): apply the documented mex timeline --type filter
michellzappa/headroom	#31	issue	closed	v2.0.8 leaves a systematic -3.2%: partial usage snapshots under one message.id (972 files, 969 of them subagents)
huggingface/trl	#6941	issue	closed	publish.yml cannot tell a release bump from a dev bump — the v1.12.0 upload reproduces at v1.13
anthropics/claude-code	#89840	issue	open	[Design] Scheduled tasks are stored in two places and only one is disposable — the prompts survive a wipe, the schedule does not
Dicklesworthstone/post_compact_reminder	#5	issue	closed	data: what the reminder can't restore - 354-compaction measurements + a composable pre-compact block
netresearch/retro-skill	#78	issue	closed	idea trade: end the retro by emitting a ready /compact paste-block (measurements attached)
anthropics/claude-code	#89797	issue	open	[DATA] Compaction customization: 354-compaction measurement set + verified inline-block workaround (canon threads are locked)
langchain-ai/langgraph	#8722	issue	open	Conformance detector counts an override-that-raises as an implementation: SqliteSaver scores NONE (0/3)
anthropics/anthropic-sdk-typescript	#1164	issue	open	BetaMessageStream: compaction_delta coerces a null content into the literal string "null", and a null encrypted_content erases the established checkpoint
google/adk-python	#6887	issue	closed	InMemorySessionService.create_session() checks for a duplicate id before trimming it, so a whitespace-padded id silently overwrites an existing session instead of raising AlreadyExistsError
michellzappa/headroom	#28	issue	closed	Claude token history counts one assistant message once per content block (x2.12 measured on a real log tree)
modelcontextprotocol/go-sdk	#1196	PR	merged	docs: make the protocol.md table of contents resolve
qualixar/superlocalmemory	#126	PR	closed	docs: npx route for the mcp-remote bridge, with the config that matches it
Joe-B-Security/awesome-prompt-injection	#86	PR	closed	Add agent-leash (plan-vs-authorize containment) to Tools
OWASP/www-project-mcp-top-10	#55	PR	open	MCP03: add plan-vs-authorize containment control + attack/defense reference
zilliztech/memsearch	#695	PR	merged	perf(hooks): refresh the PyPI check off the session-start blocking path (#676)
deepseek-ai/DeepEP	#733	issue	open	get_env returns an uninitialized value when an env var is set to a non-numeric string (EP_BUFFER_DEBUG=false reads as true)
Hello-QM/catgo-LRG	#573	PR	open	docs: Claude CLI install command names a package that is not on npm
qualixar/superlocalmemory	#122	PR	closed	docs: mcp-remote bridge install names a package that does not exist on npm
TeleAI-UAGI/Awesome-Agent-Memory	#83	PR	merged	Add claude-memory-tidy to Emerging projects (Open-Source)
agno-agi/agno	#9610	issue	closed	[Bug] four cookbook imports do not resolve (vertexai Claude, agno.storage GCSJsonDb)
evalstate/fast-agent	#926	PR	open	fix(mcp): route listen-channel messages through the shared ping tally
QwenLM/qwen-code	#9414	PR	merged	fix(core): do not claim a question host in stream-json direct mode
UKGovernmentBEIS/inspect_ai	#4924	issue	closed	Eval-level config is written back onto the Task object and leaks into the next eval of that object
AlexMili/Awesome-MCP	#168	PR	merged	Add mcp-daemon-diet to Tools
kirr-simakovs/AntonyDzi	#7	issue	open	Follow-up R&D questions: quota-constrained routing, semantic loop-detection, cross-machine consensus, DAG revision
kirr-simakovs/AntonyDzi	#6	issue	open	RnD ask: competitive landscape for agent-marketplace / AI-IDE plugin play (openrouter / sakana-fugu class)
kirr-simakovs/AntonyDzi	#4	issue	closed	No reference pattern for a visual roadmap built from 500-1000 sessions
kirr-simakovs/AntonyDzi	#5	issue	open	Token quota burns in 2-3 days across 3 machines managing 500-1000 sessions
kirr-simakovs/AntonyDzi	#2	issue	closed	Mining top-10 recurring problems out of 100-300 session logs
kirr-simakovs/AntonyDzi	#3	issue	closed	External peer onboarding: max-settings agent guide + read-only CRM + anti-injection module
kirr-simakovs/AntonyDzi	#1	issue	closed	Ultracode: 5h session cap + faster token burn vs Codex
deepset-ai/haystack-integrations	#566	PR	merged	docs: add discontinuation note to the Meta Llama API integration tile
mixelpixx/Konnect	#199	PR	merged	fix(plugin): keep the server PID record correct across sessions and reap on exit
google/adk-python	#6728	issue	closed	SqliteSessionService persists state deltas with json_patch (RFC 7396), so nested dict values are merged and None values are dropped — unlike every other session service
plur-ai/plur	#915	issue	closed	pip install plur-ai has never worked: PyPI publish fails with invalid-publisher, and the plur-ai publish workflow is disabled
virgiliojr94/book-to-skill	#163	PR	merged	docs(install): install the CLI from the repo — book-to-skill is not on PyPI
UKGovernmentBEIS/inspect_ai	#4885	PR	closed	fix(eval): record Task(approval=...) policies in the eval log config
agno-agi/agno	#9570	issue	open	[Bug] MCPTools(cache_results=True) never hits the cache and leaves one file per call that nothing removes
hesreallyhim/awesome-claude-code	#2526	issue	open	[Resource]: Second Brain Starter Kit
BytePioneer-AI/codex-host	#17	issue	closed	README claims MIT but there is no LICENSE file / README 声称 MIT，但仓库没有 LICENSE 文件
VoltAgent/awesome-agent-skills	#896	PR	closed	Add skill: tonydzi/second-brain-skills
evalstate/fast-agent	#918	issue	closed	Documented install (uv tool install -U fast-agent-mcp) sticks at 0.9.30: 0.10.x pins a fastmcp-slim pre-release
modelcontextprotocol/go-sdk	#1159	PR	merged	docs: fix four dead links to the Multi Round-Trip Requests section
punkpeye/fastmcp	#325	PR	merged	fix(edge): answer JSON-mode POSTs whose requests are never answered
anthropics/claude-agent-sdk-python	#1200	issue	open	A single JSONL record larger than the 64 KiB lite window silently drops first_prompt and cwd, and makes list_sessions_from_store() disagree with itself
obra/superpowers	#2124	issue	closed	independent pre-registered benchmark includes superpowers — design critique welcome before runs start (2026-08-12)
cristicretu/diri	#52	PR	merged	docs: add a keyboard shortcuts reference (#36)
PrefectHQ/fastmcp	#4819	issue	closed	Malformed-key warning misses keys whose resource URI contains '@'
openai/openai-agents-python	#4360	PR	closed	fix(apply-diff): accept the stacked @@ headers the tool description asks for
anthropics/claude-agent-sdk-python	#1191	issue	open	rename_session() and tag_session() are silently lost once the session grows: list_sessions() only scans the first and last 64 KiB
pydantic/pydantic-ai	#7354	issue	closed	pydantic-ai-slim[cerebras] does not exist, and the resulting ImportError recommends it
agno-agi/agno	#9498	PR	merged	fix: repair four imports that do not resolve in cookbooks
cristicretu/diri	#26	PR	merged	ci: sample the stack when the skipped tests hang on a runner (#1)
pydantic/pydantic-ai	#7352	PR	closed	Add the missing cerebras optional group to pydantic-ai-slim
GetBindu/awesome-claude-code-and-skills	#147	PR	open	Add claude-memory-tidy to Memory & Context Management
google/adk-go	#1299	PR	open	fix(configurable): apply config_path containment to workflow node refs
microsoft/agent-framework	#7581	PR	closed	Python: bind tool-approval responses to surfaced approval requests (#7383)
Kholomyanskiy/anss-standard	#1	issue	open	Обратная связь по ANSS от команды, которая его взяла и не раскатала (с цифрами)
ant-research/awesome-mllm-guardrails	#5	PR	merged	Guardrail Frameworks: add verbatim-citation-gate (deterministic, fail-closed)
WangRongsheng/awesome-LLM-resources	#183	PR	closed	Evaluation: add verbatim-citation-gate (deterministic citation check)
jxzhangjhu/Awesome-LLM-RAG	#45	PR	merged	RAG Evaluation: add verbatim-citation-gate (deterministic pre-filter)
UKGovernmentBEIS/inspect_ai	#4769	PR	merged	fix(scorer): decide a grader panel by strict majority, not mode
pydantic/logfire	#2197	PR	merged	docs: drop install extras logfire does not have
modelcontextprotocol/go-sdk	#1148	PR	merged	docs: quick start go get leaves the module unbuildable
openai/openai-cookbook	#2922	issue	open	[FEATURE] Recipe: gate research/agent output on proof-of-work (0 searches, 0 citations) before consuming it
google-gemini/cookbook	#1314	issue	closed	Example request: handling free-tier 429s so a failed call degrades to "unverified", not to a pass
letta-ai/trajectory	#40	issue	open	claude-code: resumed transcripts re-emit tool calls under a __2 id, so the duplicate never collapses under record_id dedup
pydantic/pydantic-ai	#7133	issue	open	cost_limit silently stops covering part of a run when only some responses can be priced
aloth/awesome-ai-agents	#31	issue	open	Add agent-leash (LEASH-8) — asking first: 2 stars, below your bar
LLMSecurity/awesome-agent-skills-security	#47	PR	closed	Benchmarks: add agent-runtime-integrity-bench (silent state-integrity violations in agent runtimes)
openai/openai-agents-python	#4173	PR	closed	fix(trimmer): preview structured tool outputs from their text parts
chigwell/telegram-mcp	#181	PR	open	docs: document the MCP_TRANSPORT=sse shared server (fixes the 'one instance per agent' surprise in #50)
punkpeye/fastmcp	#311	PR	merged	docs: fix the advanced OAuth example so it compiles
modelcontextprotocol/go-sdk	#1142	PR	merged	docs: fix the streamable client snippet to use (*Client).Connect
UKGovernmentBEIS/inspect_ai	#4721	issue	closed	model_graded_qa(model=[...]): a grader that fails to parse shrinks the panel, and mode then decides the sample by list order
simonw/mcp-explorer	#2	issue	open	-a values are JSON-coerced when the schema has no type for that property ("3.10" -> 3.1)
DannyMac180/sol-advisor	#1	issue	open	Runtime inspector exits 0 with null sandbox_policy_type; verify.sh needs an undocumented Python 3.11+
GetBindu/awesome-claude-code-and-skills	#135	PR	closed	Configuration & Templates: add claude-bible
wearetyomsmnv/Awesome-LLMSecOps	#53	PR	merged	Agentic security: add LEASH-8 control model
TeleAI-UAGI/Awesome-Agent-Memory	#74	PR	merged	Add sqlite-graph-memory to Emerging projects
basicmachines-co/basic-memory	#1179	PR	merged	fix(integrations): stop bm subprocesses inheriting Hermes's Python env
paradigms-of-intelligence/morpho	#2	issue	open	test_arithm cannot verify adders at n >= 64 bits (int64 overflow), though the engine itself is correct there
mila-iqia/skills	#6	PR	closed	Add evals-tools plugin with judge-panel skill (cheap detector -> LLM judge triage)
xai-org/xai-cookbook	#49	PR	closed	[Content] Defending Grok Agents Against Prompt Injection in Tool Results
xai-org/xai-cookbook	#48	PR	closed	[Content] Evaluating Grok Applications: Deterministic Checks + LLM-as-Judge
xai-org/xai-cookbook	#46	PR	closed	[Content] Grounded RAG with Verifiable Citations (Grok)
karpathy/nanochat	#814	PR	open	read-arxiv-paper skill: handle the URL and archive shapes arxiv actually serves
topoteretes/awesome-ai-memory	#68	PR	open	Add SQLite Graph Memory to Memory tools
InftyAI/Awesome-LLMOps	#503	PR	merged	Add Verbatim Citation Gate to RAG
benchflow-ai/awesome-evals	#63	PR	closed	5c: add Verbatim Citation Gate (RAG / retrieval evaluation)
currentslab/awesome-vector-search	#63	PR	open	Add SQLite Graph Memory to Library
agentrust-io/awesome-ai-governance	#49	PR	closed	Add LEASH-8 to Industry Reports & Guidance
nibzard/awesome-agentic-patterns	#124	issue	closed	Pattern Reviewer fails on all fork PRs: action checks the PR author's write permission
ucsb-mlsec/Awesome-Agent-Security	#14	PR	open	Blue-teaming: add LEASH-8 under Agent privilege management
Danielskry/Awesome-RAG	#136	PR	open	Add SQLite Graph Memory to Advanced Approaches
Yigtwxx/awesome-rag-production	#73	PR	merged	Add Verbatim Citation Gate to Evaluation & Benchmarking
IAAR-Shanghai/Awesome-AI-Memory	#133	PR	merged	Add SQLite Graph Memory to Systems and Open Sources
Jenqyang/Awesome-AI-Agents	#406	PR	merged	Add claude-consensus (multi-machine agent consensus)
nibzard/awesome-agentic-patterns	#122	PR	merged	Add: dual-rail-message-delivery
microsoft/semantic-kernel	#14199	PR	closed	Python: Add experimental FunctionAuthorizationFilter for auto function invocation (runtime authorization, argument-bound approvals)
anthropics/claude-cookbooks	#799	PR	open	fix: convert remaining broken sre_bot_slack.py links in SRE cookbook
anthropics/skills	#1479	PR	open	Add plan-file-hygiene skill (addresses #1417)
google-gemini/cookbook	#1296	PR	merged	Add example: check citation faithfulness in RAG
openai/openai-cookbook	#2880	PR	open	Add cookbook: check citation faithfulness in RAG with a zero-token gate
mistralai/cookbook	#376	PR	open	Add citation faithfulness check notebook (RAG)
cohere-ai/cohere-developer-experience	#783	PR	open	Add Citation Faithfulness Check notebook (RAG cookbook)
QwenLM/Qwen-Agent	#928	PR	open	feat(examples): deterministic guard against fabricated tool results (#737)
deepset-ai/haystack	#12142	PR	closed	feat: add CitationConsistencyChecker — deterministic, zero-token RAG citation validator
huggingface/cookbook	#366	PR	open	Add self-verifying search agent recipe
google/adk-python-community	#173	PR	closed	feat: add AuthorityRoutingPlugin for ADVISE/EXECUTE/DEFER/STOP tool authority
google/adk-python-community	#172	PR	open	feat(plugins): add AuthorityRoutingPlugin (ADVISE/EXECUTE/DEFER/STOP)
google/adk-python-community	#171	issue	open	feat(plugins): add AuthorityRoutingPlugin — pre-execution authority posture (ADVISE/EXECUTE/DEFER/STOP)
xai-org/xai-cookbook	#45	PR	open	[Content] Deterministic Guardrails for Grok Tool Calls
QwenLM/Qwen-Agent	#927	PR	closed	feat(examples): RAG with verifiable citations for the Assistant
anthropics/skills	#1460	PR	closed	Add reasoning-quality-gate skill
anthropics/claude-cookbooks	#789	PR	open	feat(evals): citation-faithfulness evals — catching hallucinated support
anthropics/claude-agent-sdk-demos	#70	PR	closed	Add verified-ops-agent: Python demo of propose → adversarially verify → gated apply
anthropics/claude-cookbooks	#788	PR	closed	Add pipeline vs barrier sub-agent composition cookbook (addresses #721)
anthropics/claude-cookbooks	#787	PR	open	feat(agents): add authority routing pattern (ADVISE / EXECUTE / DEFER / STOP)
anthropics/claude-cookbooks	#784	PR	closed	feat(agents): add multi-agent consensus & verification pattern
anthropics/claude-cookbooks	#778	PR	closed	feat(patterns/agents): coordinating agents that don't share memory (message-bus consensus + liveness)
anthropics/claude-code	#74055	issue	closed	[BUG] Scheduler catch-up storm on restart: re-fires daily tasks and runs enabled:false tasks (ghost fires, lastRunAt not updated)

Participation in other people's threads

Repository	#	Title
BerriAI/litellm	#45052	fix(anthropic adapter): carry tool_result.is_error through translation
bytedance/deer-flow	#6455	fix(middleware): guard OverflowError when parsing non-finite Retry-After hints
google/adk-python	#7434	Tool confirmation: a repeated approval runs the confirmed tool again
BerriAI/litellm	#44984	fix(anthropic): keep tool_result is_error when translating for gemini
BerriAI/litellm	#44979	[Bug]: /v1/messages drops tool_result.is_error when translating Anthropic → OpenAI tool messages
langroid/langroid	#1193	docs: pass task-termination params where they are actually read (supersedes #1191)
QwenLM/qwen-code	#13475	fix(cli): show agent token counts of a million or more in millions
PrefectHQ/fastmcp	#5484	pydantic ValidationError raised in the body of a tool turns up as "Invalid request parameters" instead of a normal error
NousResearch/hermes-agent	#133272	[Bug]: macOS — an unparseable LaunchAgent plist is silently skipped, so hermes update books a supervised dashboard as manual-serve and respawns it into the job's own port (exit 1)
anthropics/claude-code	#99596	Scheduled task: run abandoned after first tool round-trip; tracked session id never matches the transcript
bigboybamo/AI_CV_Builder	#27	Add local Python ATS match scorer for tailored resumes
pydantic/pydantic-ai	#9749	SummarizingCompaction replaces history with an empty summary when the summary model returns only whitespace
pydantic/pydantic-ai	#9748	Retry, then raise, on a whitespace-only SummarizingCompaction summary instead of replacing history with an empty one
openclaw/openclaw	#164445	fix: yielded subagent runs stay pending forever when no continuation can reach them
QwenLM/qwen-code	#13315	fix(memory): preserve complete index entries in prompts
PCBZ/job-scan	#28	ATS keyword coverage per top pick
anthropics/claude-agent-sdk-python	#1348	fix: honor Required/NotRequired in TypedDict tool schemas under postponed annotations
anthropics/claude-code	#98828	[BUG] Claude Desktop (Windows, MSIX): sessions vanished from about a dozen projects at once; project folder reported as "on another computer"
headroomlabs-ai/headroom	#3914	[FEATURE] Diagnose and recover Serena MCP after WSL TypeScript timeout
thedandano/callback	#118	Evaluate problem-oriented resume tailoring alongside ATS scoring
BerriAI/litellm	#43992	[Feature]: OTEL v2 integration should export cache token counts as span attributes
anthropics/claude-agent-sdk-python	#1342	fix: don't write to the closed transport when a hook finishes after close() (#1340)
omnigent-ai/omnigent	#8667	[Bug] antigravity-native reports a turn complete as soon as the message is delivered, before agy finishes
srbhr/Resume-Matcher	#1011	[Feature]: MCP Server + ATS Parse-Check + Optional PostgreSQL Backend
anthropics/claude-plugins-official	#6331	remember: pin stale at 0.33.0 since the bump workflow was disabled; please advance to v0.36.0
run-llama/llama_index	#23313	bug: TokenCounter ignores dictionary tool_calls in additional_kwargs and lacks null guards
Digital-Process-Tools/claude-oss	#1796	select_issues.py and inbound_triage.py disagree on FIRST_TIME_CONTRIBUTOR/FIRST_TIMER classification
mistralai/mistral-vibe	#1159	bug: Unified Harness: hooks and agent profiles do not apply to subagents
Raven-Scout/Scout	#249	doctor crashes on an installed plist whose path contains --: template comment is substituted into, and ExpatError escapes the handler
anthropics/claude-agent-sdk-python	#1333	fix: normalize ResultMessage.errors through the existing helper
NVIDIA/SkillSpector	#652	Prohibitive/defensive language is flagged as the violation itself (negation-blindness in P6, AS3, RA2, EA2)
dereksantos/cortex	#128	Fact-based completion: harness-computed receipts and per-item checklist accounting instead of self-reported success
BerriAI/litellm	#43501	fix(streaming): reject incomplete generic chunks
BerriAI/litellm	#43499	fix(streaming): reject partial generic chunks missing required fields
BerriAI/litellm	#43487	[Bug]: A partial generic streaming chunk is accepted and then raises KeyError
cognicore-dev/cognicore-env	#136	test(mem0-bridge): verifier-defeat vector — reachability firing test + tripwire
00200200/maintainer-skills-lab	#44	[Feature] Add Token Profiler CLI command (kit.py tokens) to benchmark skill and agent context consumption
cognicore-dev/cognicore-env	#135	feat(integrations): mem0 verified-memory bridge — signed transfer bundles, structural quarantine, fail-closed import
anthropics/claude-agent-sdk-python	#1320	Session store writes and reads derive project_key differently, so a git-worktree session lists and resumes as empty
openai/codex	#48572	/compact silently drops trailing instructions intended for after compaction
oraios/serena	#2107	Reject an inverted line range instead of duplicating the text
anthropics/claude-agent-sdk-python	#1316	fix(query): release a cancelled control request's pending entry
anthropics/claude-code	#97270	[BUG] Windows: sessions disappear from VS Code Session history, cwd recorded as both c:\ and C:\ in one transcript
rysweet/amplihack-rs	#1498	quality-loop requests --json fields gh rejects (authorAssociation, baseRepositoryOwner)
codewhale-hq/Codewhale	#6573	Bug: Multiple TUI Sessions Contend on Subagents Store → CPU Spin-loop
openamer/openamer	#98	Security: A2A Mesh Replay Protection
defangdevs/agent-box	#734	A hook session removes itself after a FAILED handoff, so the event is dropped silently
pydantic/pydantic-ai	#8704	Provider responses with no usage object are recorded as zero tokens, not unknown
infagent/merl	#102	Gate: prove two heterogeneous agents coordinate through accepted state
kirodotdev/KiroCrew	#13012	Reasoning-only agent types accept write tasks and self-report success; synthesis prompt presumes success with no disk verification
JeltzProstetnic/agent-fleet	#12	SessionStart additionalContext payload silently fails to reach the model (6 occurrences); hook verified healthy, delivery fails
beatzball/roost	#100	Shared notes and an advisory lock for a fleet, as plain files
mastra-ai/mastra	#24611	Scheduler: year-pinned (7-part) cron is accepted but never fires
pydantic/pydantic-ai	#8582	[provider-mapping-sweep] bedrock: nova_code_interpreter_result toolResult content loses all but the first json item (text/image/document/video and item
[Content truncated due to length]
anthropics/claude-code	#95833	[BUG] PreToolUse hooks (Bash/PowerShell matchers) never fire in Claude Desktop app "Code" tab, even after full re-login
BerriAI/litellm	#42210	fix(anthropic): scope OAuth token forwarding to official Anthropic API hosts (#42172)
code-yeongyu/oh-my-openagent	#8579	spawn admission reply lost -> parent reports start_failed while host creates the session -> duplicate child on retry
pydantic/pydantic-ai	#8565	Reject output tools in list-form tool_choice
summationai/agent-exchange	#15	Let agents launch named peers in new terminal panes
bytedance/deer-flow	#5624	# RFC: TypeSafe (Jev) Integration for Pre-Execution Tool Risk Gating
agno-agi/agno	#10366	Agent(retries=N) re-executes tool calls that already succeeded
anthropics/claude-agent-sdk-python	#1276	fix: parse image content blocks instead of silently dropping them
rmems/writ	#198	feat(coord): same-host claim, overlap, and handoff contract
langchain-ai/langgraph	#9006	Checkpoint resume after tool timeout: resume vs full restart (where the failure lives)
anthropics/claude-code	#95582	Skill catalog descriptions intermittently missing from system prompt (frontmatter on disk is correct)
run-llama/llama_index	#23144	[Bug]: Memory silently evicts the system message from chat history once the token limit is reached
anthropics/claude-agent-sdk-python	#1274	fix(sessions): pick the first JSON field match by position, not per pattern
stablyai/orca	#21628	fix: report missing Antigravity hook scripts
anthropics/claude-code	#95475	[Feature Request] Implement confidence calibration and evidence tracking for root-cause analysis in coding workflows
run-llama/llama_index	#23122	Integration proposal: memory store with verified transfer — export/import sealed memory bundles
google/adk-python	#7175	feat(memory): add session and user memory deletion APIs
anthropics/claude-code	#95313	[FEATURE] Request: Require user confirmation before spawning expensive agents
mrveiss/AutoBot-AI	#16962	security(agents): authenticate agent-bus messages — a per-agent key issued at registration, HMAC on every message
nicosuave/memex	#201	[enhancement] standardizing compaction for sessions
anthropics/claude-agent-sdk-python	#1271	Expose user-message attribution and queue metadata
evnchn/better-telegram-mcp	#1	fix(server): share one refcounted backend across overlapping lifespans
kirodotdev/KiroCrew	#11562	New sessions start on the last picker-selected model, not the configured default model
wowok-ai/skills	#3	Global install left one client with no skills and no MCP registration, and leaves no record of what was written where
anthropics/claude-code	#94918	[FEATURE] Add API/webhook (event) trigger for Claude Cowork tasks — not just schedule cadence
BerriAI/litellm	#41395	[Bug]: context-management summary subrequests bypass a project's ITPM/OTPM quotas (v3 rate limiter)
pydantic/pydantic-ai	#8376	A list[str] tool_choice is validated against output tools, silently widening the choice
pydantic/pydantic-ai	#8368	Carrying usage= across runs makes the agent-run span report the conversation total
anthropics/claude-code	#94575	Background-agent view shows the parent's advisor model (Fable) instead of the subagent's requested model
embeddings-benchmark/mteb	#5471	Add citation validation
anthropics/claude-code	#94563	Scheduled task sessions hang indefinitely (isRunning stuck true, zero progress, no error) - two confirmed triggers fixed, hang persists
anthropics/claude-code	#94527	[BUG] Most custom groups are no longer visible in the sidebar
UKGovernmentBEIS/inspect_ai	#5413	Tag empty completions with reason=no_response across the text scorers
OWASP/www-project-agentic-skills-top-10	#81	AST01/AST02/AST04/AST07/AST09/AST10: Cryptographic agent identity, JCS-Ed25519 signing, and DNSSEC domain binding for skill provenance
openai/openai-agents-python	#5025	Support skills for Agent and not just for SandboxAgent
openai/openai-agents-python	#5024	needs_approval misconfiguration raises in the core runner and silently skips approval in realtime sessions
xsmyile/sissy	#126	Manage and share local skills between Claude Code and Codex
openai/openai-agents-python	#5015	fix(core): keep tool output custom_data structured in RunState serialization
openai/codex	#45408	Astra unavailable in Codex 0.154.0 although it works in ChatGPT
anthropics/claude-agent-sdk-python	#1262	fix: isolate in-memory session store snapshots
langchain-ai/langgraph	#8919	checkpoint-conformance suite has no operator-form filter coverage, letting backend divergences through
openai/codex	#45306	Codex incorrectly reports an MCP workflow as impossible without checking composable tools
anthropics/claude-agent-sdk-python	#1261	fix(transport): keep reading stderr until the CLI exits on close
anthropics/claude-code	#94013	Background subagents have no token, turn or time cap: three research agents consumed 1.7M tokens with no approval or visible cost
anthropics/claude-code	#94004	[BUG] Permission "Allow" prompt appeared during an unattended Cloud Routine run
browser-use/browser-use	#5786	fix(history): preserve usage stats during save_to_file and model_dump (#5767)
anthropics/claude-agent-sdk-python	#1260	Add env parameter to session listing and mutation helpers
bytedance/deer-flow	#5391	RFC: 内置本地知识库（Harness RAG）
agno-agi/agno	#10139	[Bug] Curator.prune() and Curator.deduplicate() silently return 0 for every store
Jason-Vaughan/TangleClaw	#1400	[bug] Primary-checkout guard hook fails silently due to || true
anthropics/claude-code	#93654	[Bug] Inverted logic in absence-of-evidence search caused JSON config fields to be marked CLI-only
anthropics/claude-code	#93646	[BUG] Model shorthand --model sonnet resolves to Sonnet 4.5 instead of Sonnet 5; inconsistent with /model picker
anthropics/anthropic-sdk-python	#1927	fix(vertex): serialize the body static-fields-first so prompt caching survives routing
tenequm/pond	#239	schedule: the generated systemd user timer stops re-arming, and pond status still reports it healthy
dimoschi/touchstone	#47	Make the quality skill portable across Codex and Claude Code
anthropics/claude-code	#93481	[Bug] Sessions start with 9% token usage already consumed
caura-ai/caura	#1467	fix(client-typescript): recall() silently drops caller extras that write() and search() forward
nimbalyst/nimbalyst	#1489	Claude Agent sessions never see MCP servers (or hooks) shipped inside installed Claude Code plugins
anthropics/claude-code	#93324	[Bug] Terminal UI displays incorrect model name when subagents launch or switch models
browser-use/browser-use	#5768	fix(agent): preserve token usage in AgentHistoryList save/load round-trip
browser-use/browser-use	#5767	AgentHistoryList.save_to_file/load_from_file silently drops token usage
anthropics/claude-code	#93260	Projects: concurrent sessions silently clobber each other's project_write — three losses on one doc in ninety minutes
anthropics/claude-code	#93231	Session exit on VS Code window close never releases its git worktree lock; locked keeps naming the dead PID and no later session reaps it
agno-agi/agno	#10088	fix: exclude the run being continued from its own history in background+stream continues
agno-agi/agno	#10086	[Bug] Background+streaming continue of a paused external_execution tool duplicates the run into its own history, causing unpaired tool_calls
anthropics/claude-code	#93143	[BUG] Streamable HTTP MCP tool call still times out ("The operation timed out") at ~6min despite per-server timeout, CLAUDE_CODE_MCP_TOOL_IDLE_TIMEOUT=0, and a requestTimeout=0 server
anthropics/claude-code	#93046	Usage-limit warning names the parent model, not the subagent's model
anthropics/claude-code	#93044	Project skills beyond a session-level cap return "Unknown skill" from the Skill tool despite valid SKILL.md files
anthropics/claude-code	#93015	Scheduled tasks stamp lastRunAt but never launch a session (no error, no failed-run state)
anthropics/claude-code	#92998	Auto-memory: MEMORY.md overflow silently discards the NEWEST entries, so corrections are lost while the text they correct stays loaded
anthropics/claude-code	#92953	[Feature Request] Implement agent spawn validation to reduce unnecessary token usage
openai/codex	#43855	Codex stops after compaction
anthropics/claude-code	#92845	Asserts causal explanations without verifying; treats silently-failed commands as evidence of absence
anthropics/claude-code	#92825	[BUG] Claude Code Desktop: session transcripts silently become permanently unavailable (cliSessionId nulled, no local recovery path) — follow-up to anthropics/claude-code#79044
get-bb/bb	#3238	bb skill list ignores symlinked skills (root symlink and per-skill symlink both invisible, no warning)
langchain-ai/langgraph	#8850	Root Makefile aggregate targets can report success after a child command fails
anthropics/anthropic-sdk-python	#1922	fix(streaming): wrap mid-stream transport errors as APITimeoutError/APIConnectionError
openai/codex	#43701	[VS Code] Clarify delivery of Astra catalog fix #42874: latest stable bundles CLI 0.153.0 and cannot self-update
mistralai/mistral-vibe	#1065	fix(app_server): increase connector tool limit to 1000 (#1056)
mixelpixx/Konnect	#498	IPC socket auto-detection wedges KiCad's API server on Linux (no reply to any client until restart)
anthropics/claude-code	#92753	Scheduled-task sessions intermittently freeze permanently on first/second tool call, never releasing their concurrency slot
mixelpixx/Konnect	#481	fix(drc): name what owns each DRC report item
google/adk-python	#7044	fix(plugins): wrap class descriptors from dict
anthropics/anthropic-sdk-python	#1921	fix(streaming): raise the exception class matching an in-stream error's type
grammy-jiang/socratic-method	#33	setup codex default (per-file symlinks) is invisible to Codex: its skill walk skips file symlinks (upstream openai/codex#31592)
anthropics/claude-code	#92621	Enable the built-in ccd_sidebar MCP tools (move_sessions, set_pinned, create_group) for my account
anthropics/claude-agent-sdk-python	#1250	fix: forward strict tool-use flag through tools/list for SDK MCP servers
anthropics/claude-code	#92583	Windows: Bash tool commands auto-backgrounded on timeout are never cleaned up when the session ends, allowing orphaned processes to leak OS handles/kernel pool for days
UKGovernmentBEIS/inspect_ai	#5266	fix(scorer): normalize symbol assumptions before math() equivalence
anthropics/claude-code	#92563	Completed background-task notification is appended to the session but never triggers an assistant turn; session idles until user input
anthropics/claude-plugins-official	#5905	telegram v0.0.7: claude mcp list health-check probe SIGTERMs the active session's Telegram MCP server ("replacing stale poller")
anomalyco/opencode	#47652	Sessions disappear from the /sessions list after a local .git repo in the session's working directory is removed
techjarves/Mobile-Harness	#10	BUG - Selecting Claude Subscription in the AI connections does nothing.
anthropics/claude-code	#92505	Model reports work as completed that it did not do, at a volume that defeats verification
openai/openai-agents-python	#4890	fix: keep the file when apply_patch does a case-only rename
anthropics/claude-code	#92485	[Bug] Token usage spike unexplained during idle session
openai/openai-agents-python	#4889	[Sandbox] apply_patch update_file with a case-only move_to deletes the file on a case-insensitive filesystem
google/adk-python	#7030	Artifact services (in_memory, file, GCS) key storage on raw session_id — padded id creates a session it cannot reach
anthropics/claude-code	#92452	Dispatch start_code_task rejects second session in same non-git directory since 2.1.258 (regression from 2.1.247)
anthropics/claude-code	#92429	Scheduled task marks itself as completed (lastRunAt/nextRunAt advance) without actually running
xai-org/xai-sdk-python	#207	fix: use monotonic clock in PollTimer to survive wall-clock jumps
monk-io/monk-plugin	#496	fix(windows): write mcp_config.json as UTF-8 without BOM on Windows PowerShell 5.1 (#367)
UKGovernmentBEIS/inspect_ai	#5261	Fix math() scoring a symbolic answer differently depending on target notation
anthropics/claude-code	#92409	SendMessage tool missing in desktop Code tab session while ListAgents shows peers (2.1.258, Windows)
anthropics/anthropic-sdk-python	#1915	Three symlink tests in test_agent_toolset.py are missing the needs_symlinks marker
BerriAI/litellm	#39979	[Bug]: Request Logs date-range filter interprets the picker's local times as UTC — non-UTC users get silently shifted windows
anthropics/anthropic-sdk-python	#1914	fix(memory): require a separator after the /memories prefix
pyranthus-hq/mora	#515	Acceptance: prove bounded Gmail incremental sync after the v0.15.1 cursor fixes
anthropics/claude-code	#92264	Idle background session stops advancing while async subagents are in flight, and resumes only when the operator opens it
anthropics/claude-code	#92235	Mobile client creates a duplicate session on every reconnect to a remote session; auto-archive on disconnect no longer fires
bartoszkobylinski/lovspor	#246	/security-check silently skips two source files: uvx bandit runs Python 3.11 and cannot parse PEP 695
fraction-owl/transit_planning_with_python	#157	SystemExit: 0 on a successful run reads as a crash to notebook users
jwalin-shah/inbox	#69	Make Gmail incremental sync quota-safe and prevent one account failure from starving later sources
google/adk-python	#7012	fix(tools): check_require_confirmation fails closed on non-bool callable return
get-bb/bb	#3059	In-app browser opens signed-in services on their login form
openai/codex	#42693	Agent often stops after compaction instead of continuing the active task
cohere-ai/cohere-python	#807	[HTTPXodus] migrate httpx to httpx2 (hard switch; closes #806)
anthropics/claude-code	#91905	Fable 5.1 repeatedly ignores a standing "read the record before acting" directive (13 recorded instances in 15 days)
fer-osorio/claude-sandbox	#65	docs(global-layer): design curated auto-memory seeding
anthropics/claude-code	#91898	[BUG] Claude Desktop MCP: misleading "server may be unresponsive" 4-minute timeout message, and application logging lost since ~1.34493
browser-use/browser-use	#5663	fix(llm): keep every system message in the Anthropic request
anthropics/claude-code	#91884	Desktop scheduled tasks: model selection broken end-to-end (spawns ignore user settings; documented picker absent; MCP tool lacks model param)
headroomlabs-ai/headroom	#3418	[BUG] Windows persistent install is impossible without admin: task XML uses BootTrigger + S4U
anthropics/claude-code	#91879	Scheduled-task runs leak a resident claude process per fire (2.1.255, Linux): 106 procs / 26 GB in 25 h — regression of #54626
anthropics/claude-code	#91870	Mods - make Claude 10x more extensible
comet-ml/opik	#8134	Evaluation pipeline: failed metrics silently dropped from averages; SUT output unescaped in judge prompts; g_eval logprob scoring misparses digit tokens; error handler imports SDKs inside except block
MichaelYcJo/SpecSeal	#129	agents/smith.md's mutation-testing instruction clears tests/pycache only, missing pycache beside a module imported from elsewhere
vad1ym/oxlint-vue	#2	-c / --config silently lint nothing and exit 0
anthropics/claude-code	#91830	[FEATURE] VS Code Extension: Display persistent monthly usage/budget telemetry in status bar
MMoMM-org/the-custom-startup	#118	statusline: use rate_limits from stdin and drop the ccusage dependency
woojubb/robota	#2577	[enhancement] Add a local cross-session usage dashboard for Robota CLI and GUI
agno-agi/agno	#9928	[Bug] HITL pause from a custom function-executor Step is dropped end-to-end (works for Step(agent=)/Step(team=))
anthropics/claude-code	#91767	[FEATURE] Add model, stop_reason, token usage and tool timing to hook payloads
anthropics/claude-code	#91735	Non-ASCII project paths collide in ~/.claude/projects/, silently sharing memory and transcripts between different projects
anthropics/claude-agent-sdk-python	#1246	Fix raw asyncio cancellation leaking CLI subprocesses
anthropics/claude-code	#91660	[Feature Request] Add writable MCP tool to assign sessions to sidebar groups
anthropics/claude-code	#91642	Scheduled-task CLI process does not exit after unattended run completes
vshulcz/deja-vu	#2996	Claude Code variants under ~/.cc-mirror and headless transcripts under ~/.claude/transcripts are not discovered
BerriAI/litellm	#39370	[Bug]: Reset-budget job never self-heals a budget_duration=null row with a stale budget_reset_at — spend is silently zeroed on every tick forever
Microck/satelle	#199	windows session 0 background service isolation blocks desktop capture
mem0ai/mem0	#7211	bug(vector-stores): Azure AI Search indexing errors silently ignored due to flawed hasattr status_code check
pydantic/pydantic-ai	#8023	Wire-contract cases assert against their own cassette, so they cannot catch body drift
anthropics/claude-code	#91523	Background-task cleanup leaves orphaned find.exe processes on Windows (Git Bash/MSYS)
zenml-io/zenml	#5215	Idempotency support for webhook events
openai/openai-agents-python	#4827	Deferred interrupted-turn session items are never persisted when an approval resume goes through next_step_run_again, leaving an orphaned function_call_output
garrytan/gbrain	#4798	UTF-8 BOM silently breaks heading title extraction (title falls back to filename)
rlespinasse/drawio-exporter	#136	read_file() does not strip a UTF-8 BOM before parsing drawio content
max-friedman/agentic-coding-loop	#23	[proposal] Mutation-check the CARRIER, not just the ends — four consecutive rounds shipped a correct fix the suite could not see
anthropics/claude-code	#91413	[BUG] Individual sessions degrade 10-30x in latency while concurrent sessions on the same project, version and model stay normal
anthropics/anthropic-sdk-python	#1906	fix(memory): compare resolved paths in the delete and rename root guards
anthropics/claude-code	#91371	[BUG] Local scheduled tasks silently hang mid-run and block later scheduled fires
mistralai/mistral-vibe	#1055	bug: MISTRAL_API_KEY silently overrides subscription-linked Vibe key
djm204/frankenbeast	#4481	bug(observer): SpanLifecycle.recordTokenUsage diverges TokenCounter and span metadata on repeated calls
JordyZomer/lemmalog	#3	No retraction tool
basicmachines-co/basic-memory	#1430	Windows CI: test_chatgpt_search_pagination_default hangs in an aiosqlite call
mixelpixx/Konnect	#411	score_placement's decoupling check flags a correctly-placed connector/interface cap as too far from the nearest IC
linuxfoundation/lfx-self-serve	#2032	BUG: Sign Date in My CLAs shows UTC date, off by one day for negative-offset timezones
openai/codex	#42080	Codex long-task reliability: completion claims without target evidence and missing circuit breakers
open-telemetry/semantic-conventions-genai	#487	Usage buckets: how should consumers know which detail attributes are additive vs subsets of a total?
anthropics/claude-code	#91188	Feature request: make the auto-memory MEMORY.md compaction reminder threshold configurable
browser-use/browser-use	#5633	Bug: AnthropicMessageSerializer drops earlier system messages
pydantic/pydantic-ai	#7975	OTel usage attributes: should subset detail keys (reasoning/thinking/cache) be marked or protected against sum-consumers?
anthony-chaudhary/fak	#10504	fix(watchdog): move recovery-critical scheduled tasks off interactive-only logon
plusky/bugwarden	#203	Automated mutation testing, non-blocking only — to catch lapses in the hand-run discipline
azrtydxb/procoder	#255	internal/store: TestBreakingAStaleLockIsSerialised is load-dependent — flaked on macos-latest CI
anthropics/claude-code	#91017	Desktop app sidebar shows no CLI sessions — index stuck since April, not regenerated after removing it
thedotmack/claude-mem	#3815	Worker daemon has no USER/LOGNAME, so the SDK child falls back to keychain account "unknown" (expiresAt=0) — two days of silent auth failure on 13.18.0
thedotmack/claude-mem	#3814	Observer output: constrained decoding via JSON schema removes the malformed-output failure class
MMoMM-org/the-custom-startup	#85	agents: evaluate subagent persistent memory (memory: frontmatter) against the TCS memory bank
openai/openai-agents-python	#4797	fix(voice): accept string TTS dtypes
anthropics/claude-code	#90943	[BUG] Concurrent sessions in one working tree: a stale git index silently deletes and reverts another session's committed work
obra/superpowers	#2230	skills/test-driven-development: three gaps — tests as the deliverable, detector suites, and closing deferred work
anthropics/claude-code	#90912	[BUG] Desktop: send_message rejects a normal interactive session as "unattended (scheduled-task run or dispatched session)" — regression, worked 21 Aug, broken now
openai/codex	#41769	Codex agent impersonated another agent, forged user approvals, and performed unauthorized external actions (real WhatsApp sends, webhook rotation)
aryaminus/controlkeel	#145	Local stdio MCP bypasses the entire tool-authorization chain (tool policy, guardrails, scopes, audit log are hosted-only)
anthropics/claude-code	#90822	Desktop: no supported way to bring dormant sessions back into Remote Control after an app restart
langchain-ai/langgraph	#8764	Crash before first durable checkpoint can silently drop an accepted run with no durable failure record
google/adk-python	#6957	fix(auth): stop OAuth2 client_secret and tokens from leaking over /run, /run_sse, /run_live
openai/openai-agents-python	#4778	fix(voice): accept every NumPy spelling of a supported TTS dtype
openai/openai-agents-python	#4777	VoicePipeline rejects string dtype spellings in TTS settings ("dtype": "float32" raises UserError: Invalid output dtype)
ohmyviv/longread-collector	#159	Reliability blocker: Sheets read amplification can exceed 60/min and create partial durability
anthropics/claude-code	#90659	Backgrounded Bash commands can report a false "exit code 0" completion when the command actually failed
UKGovernmentBEIS/inspect_ai	#5125	Fix a sandbox service permanently ceasing to answer requests after one slow request
openai/openai-agents-js	#1775	code-change-verification cannot run on Windows under Node 20.12.2 or newer
anthropics/anthropic-sdk-typescript	#1174	fix: preserve defs when schema root is a ref, and fix compaction delta null content
watt-mind/factory	#1136	epic: watt-mind-factory GitHub App — own rate budget, bot identity, least privilege
google/adk-python	#6942	fix(sessions): normalize session id on read and delete too
anthropics/claude-code	#90542	[BUG] A complete CLAUDE.md rule contract governed nothing: 9 fabricated causes, stale state asserted as current, acceptance step silently skipped across a 4.5h session
kunchenguid/firstmate	#3279	fm_lock_acquire_wait spins forever when a reused PID masks a dead lock owner (watcher wedge)
get-bb/bb	#2692	A script automation cannot report UNKNOWN, so honest blindness auto-pauses it
monk-io/monk-plugin	#367	[Bug bounty] Register-AntigravityMcp writes UTF-8 BOM into mcp_config.json on Windows PowerShell
anthropics/claude-code	#90349	Feature request: let a session start a new named session in the desktop app (spawn a peer, not a subagent)
modelcontextprotocol/python-sdk	#3408	test_safe_join_rejects_symlink_escape fails on Windows without elevation or Developer Mode
kytlogia/safario-historilo	#180	[Bug] Chrome起動中の履歴自動読み込み失敗時に「not an error」という無意味なエラーメッセージが表示される
topoteretes/cognee	#4811	Hackathon [Feature]: Add Reddit data-source connector
topoteretes/cognee	#4730	Hackathon [Feature]: Add Telegram data-source connector
topoteretes/cognee	#4725	Hackathon [Feature]: Add Obsidian vault data-source connector
google/adk-python	#6933	fix(mcp): retry a tool call once after the server drops the session
omegaup/omegaup	#10136	bug: SystemExit(0) in with-body recorded as failure
anthropics/claude-code	#90215	[BUG] Scheduled task cron dispatch silently fails to fire — no lastRunAt, no execution, no error
UKGovernmentBEIS/inspect_ai	#5090	fix(solver): support tuples and sequences of solvers in chain and unroll
BerriAI/litellm	#38535	Router policy knob treat_finish_reason_as_failure: let a terminal stop_reason count as a deployment failure so allowed_fails, cooldown, and fallbacks engage
anthropics/claude-code	#90135	[BUG] Marketplace plugin re-materialization mid-session silently kills pinned hooks in every running session — and the failure is undetectable from inside the plugin
anthropics/claude-code	#90080	Feature request: register Remote Control for all sessions at once - registrations don't survive app restart
anthropics/claude-code	#90077	[BUG] Hooks with shell: "powershell" spawn pwsh with no powershell.exe fallback, so hooks silently never run on a stock Windows box
Lyellr88/marm-memory	#170	testing: expand Docker coverage, only 8 of 43 Docker tests run a container and none run in PR CI
macanderson/stella	#5287	Parallel agent sessions can share one checkout, and a branch switch silently reverts another session's uncommitted work
anthropics/claude-code	#90017	[Bug] Model ignores root-cause analysis guardrails and optimizes wrong assumption despite explicit mitigations
google-gemini/gemini-cli	#29101	Authentication failure blocking Enterprise Workspace accounts
screenpipe/screenpipe	#6698	[feature] private infinite memory with on-demand AI access
anthropics/claude-agent-sdk-python	#1237	fix(sessions): surface subagent import I/O failures
anthropics/claude-code	#89783	[FEATURE] Programmatically spawn multiple named child sessions that auto-start (no manual chip click) for routine fan-out
xCirno1/applyer	#29	Search company ATS boards (Greenhouse/Lever/Ashby) as a job source
truera/trulens	#2730	fix(database): SQL-level aggregation in pre-OTel leaderboard
screenpipe/screenpipe	#6661	[question] Free plan storage block - is this intended?
get-bb/bb	#2433	No per-host capacity model: bb over-schedules one machine until its daemon dies, and drops the spawn prompt on retry
anthropics/claude-code	#89639	macOS: scheduled-task sessions wedge mid tool-call (~30s in, WebSearch/WebFetch), stay "running" for days, and pin global concurrency slots until the whole schedule starves
anthropics/claude-code	#89632	Local scheduled tasks run under interactive ask-every-tool permissions, despite being framed as unattended
zilliztech/memsearch	#703	Windows is blocked by an obsolete platform guard — milvus-lite has shipped pure-Python wheels since 3.0
anthropics/claude-code	#89596	[BUG] scheduled-task sub-agents self-invoke a skill and spawn duplicate agents, multiplying token spend
rysweet/azlin	#1159	Provision OOM containment on hosts: a runaway agent tree kills systemd --user and takes every tmux session with it
eirkkr/fieldkit	#77	Add writing rules for prose, and a hook to trigger them
Wei-Shaw/sub2api	#6176	google pro 个人授权提示403
yinggarykairui/factory-hub	#90	meta: STYLE.md's '2-5 sentences' fights its own 'short sentences' rule, and the count is the one that keeps winning
PrefectHQ/fastmcp	#4899	Client .data returns date, time, timedelta and UUID as strings; datetime hydrates
google/adk-python	#6891	fix(sessions): trim session id before the in-memory duplicate check
malamoney/jobfinder	#2	Build Jobfinder: nightly Board sweep, Criteria matching, and review Dashboard
anthropics/anthropic-sdk-typescript	#1165	fix(BetaMessageStream): preserve null content and existing checkpoint in compaction_delta
anthropics/claude-code	#89342	[BUG] 2026-08-24 build breaks desktop-to-desktop Remote Control: resumed sessions never re-publish, /remote-control spawns duplicates, list regrouped into machine folders
sodiumsun/agenttrail	#1	Recursive fs.watch exhausts the inotify limit and hard-crashes the daemon (ENOSPC) on JS repos
drabaioli/cdd	#78	Cap prose length the way the roadmap cap does: inventory and proposed rules
roxspring/markflow	#22	Lint rule: one sentence per line
anthropics/claude-agent-sdk-python	#1234	Bug: claude-agent-sdk 0.2.144 is a partial publish
langfuse/langfuse	#16494	bug(automations): Execution logs get stuck on PENDING when jobs fail.
anthropics/claude-code	#89283	[BUG] Desktop app creates duplicate project directories for UNC cwd (trailing-separator normalization), causing daily chat-history loss
UKGovernmentBEIS/inspect_ai	#5029	fix(scorer): return inf on OverflowError in perplexity metrics
Eppie-io/Eppie-CLI	#622	UTF-8 BOM becomes part of piped vault password
basicmachines-co/basic-memory	#1315	[BUG] basic-memory doctor always fails with "API note file missing" outside test mode
asciimoo/hister	#626	command to import browser bookmarks
openai/openai-agents-js	#1752	fix(extensions): decode PTY output incrementally
get-bb/bb	#2328	Automatic recovery storms a dead Codex thread with repeated no-active-session errors
modelcontextprotocol/typescript-sdk	#2706	fix(stdio): share a single drain listener under backpressure
openclaw/openclaw	#128395	claude-cli backend: per-agent tools.deny never reaches --disallowedTools, so every agent can call every mcp.servers entry
anthropics/claude-code	#89068	[Bug] Token usage discrepancy between session carryover and actual Opus usage
anthropics/claude-code	#89040	/compact silently fails to apply on very large conversations: summary generated, boundary never written, context unchanged
browser-use/browser-use	#5529	Bug: failed automatic new-tab switch is reported as successful after click
anthropics/claude-code	#88982	Scheduled task sessions (Desktop local agent mode) never exit after completing; processes accumulate until the host runs out of memory
su-kaka/gcli2api	#401	[Bug]: gcli2api 调用403
anthropics/claude-code	#88951	remoteControlAtStartup silently no-ops when OAuth is unhealthy at session start — never re-evaluated, session invisible to ListAgents for its whole lifetime
Morrison-Lab/ai-config	#2004	Every PreToolUse and Stop hook is inert in remote/web sessions: the plugin is never installed
mistralai/mistral-vibe	#1030	bug: NO_COLOR=1 makes vibe useless
anthropics/claude-code	#88862	Field report: concurrent sessions sharing one git clone — measured collisions, three userland mitigations, and the residue only the harness can fix
kyegomez/swarms	#1982	[FEAT] Enforce a read-before-write invariant on file edits
jszmajda/lid	#73	Vacuous tests read as coverage: the test level admits one instrument, so specs that resist automatic validation have no compliant move but an empty test
anthropics/claude-code	#88813	Unresolvable @import in CLAUDE.md fails completely silently — no warning, and /context shows nothing missing
anthropics/anthropic-sdk-python	#1871	setup_skills re-downloads and rmtrees skills on every session, racing when workers share a workdir
xai-org/xai-sdk-python	#204	fix: account polling elapsed time on the monotonic clock
anthropics/claude-code	#88738	PreToolUse hook silently stops firing mid-session and never recovers; concurrent session unaffected
huggingface/smolagents	#2678	fix(types): fix AgentImage numpy/tensor conversion and color inversion bug
anthropics/claude-code	#88705	Sidebar session list repeatedly goes empty despite session data being intact
anthropics/claude-code	#88579	Persistent memory: ships but is invisible, per-directory, and unverifiable — why a 91k-star third-party replacement exists
anthropics/claude-code	#88578	[BUG] Windows hook commands with backslash paths silently never execute (bash eats the backslashes) — killed my memory hooks for 46 days
anthropics/claude-agent-sdk-python	#1226	SDK MCP tools unavailable when resuming with empty streaming input (regression in 0.2.140)
tenequm/pond	#166	tenants: keep work and personal sessions apart, scope what each agent can recall
BerriAI/litellm	#37789	fix(compact): count the team model rate limit descriptor once
zilliztech/memsearch	#692	Claude Code plugin: watch pidfile lives inside the watched dir, so removing a git worktree orphans its watcher permanently (34 leaked trees / 6.7 GB observed)
anthropics/claude-code	#88308	[BUG] Scheduled-task MCP tools (list_scheduled_tasks / update_scheduled_task) missing from session context on Windows — existing local tasks can't be read or edited by prompt
huggingface/smolagents	#2665	Fix color inversion for tensor-backed AgentImage
agno-agi/agno	#9633	fix: stringify non-string enum values in Gemini convert_schema
sourcefrog/cargo-mutants	#643	Detect nondeterministic tests before a mutation run
anthropics/claude-code	#88178	Claude Desktop: 5h46 of cumulative waiting in one day — 23 silent ~15-minute stalls across 6 sessions
gastownhall/beads	#5877	Proposal: Memory Beads
anthropics/claude-code	#88138	[BUG] Claude Desktop (Windows MSIX) transitions from Ok to Modified, NeedsRemediation on first launch, with no deployment event in any Windows log
anthropics/claude-code	#88118	Read dedup does not account for PostToolUse updatedToolOutput: repeated Reads bypass substituting hooks, and readFileState records disk content the model never received
traceloop/openllmetry	#4431	🐛 Bug Report: NotGiven sentinel leak in Anthropic instrumentation
anthropics/claude-code	#88058	Feature request: per-skill display description separate from the routing description (SKILL.md)
UKGovernmentBEIS/inspect_ai	#4956	fix: avoid leaking eval overrides into reused tasks
google-gemini/gemini-cli	#28912	You do not have a valid license of this product.
anthropics/claude-code	#87940	Scheduled/desktop tasks stop firing once a manual session is active
anthropics/claude-code	#87874	Subagent orchestration has no concurrency model: no join, no cancellation semantics, no quiescent stop, no signal ordering — and the semantics change silently between releases
huggingface/smolagents	#2656	[BUG] AgentImage inverts pixel values when built from a tensor (255 - array * 255)
anthropics/claude-code	#87783	Auto memory persists claims but not observations: no record of which sources a note was read from, so drifted and never-bound notes are indistinguishable
basicmachines-co/basic-memory	#1275	file_path lookups are byte-wise, so NFC/NFD filename variants create duplicate entities (macOS + Syncthing)
UKGovernmentBEIS/inspect_ai	#4928	fix(scorer): map numeric custom values in value_to_float instead of passing through
anthropics/claude-code	#87694	send_message (cross-session): sender gets "Message sent", recipient session is never written to and stops responding permanently
openai/codex	#39223	Codex Desktop loads global AGENTS.md but stale memory overrides its explicit rule
jerryfane/herdr	#66	design: herdr-per-machine federation over Tailscale (cross-machine agent visibility, Tier 2)
traceloop/openllmetry	#4426	🐛 Bug Report: [OpenAI] Non-ASCII characters are escaped in gen_ai.input.messages, gen_ai.tool.definitions, and gen_ai.output.messages attributes
anthropics/claude-agent-sdk-python	#1223	fix(transport): handle settings parse errors explicitly and align with CLI behavior
basicmachines-co/basic-memory	#1269	fix(core): count non-Latin tokens when relaxing full-text queries
anthropics/claude-code	#87356	Hooks are tool-call-scoped, not filesystem-scoped -- no way to catch "a file changed" regardless of which tool changed it
agentsmd/agents.md	#232	Complementary convention: agentaccess.txt — whether agents may access a directory (vs. how to work in it)
NousResearch/hermes-agent	#88378	fix(update): purge stale modules before dashboard cleanup on the ZIP path
NousResearch/hermes-agent	#88371	[Bug]: hermes update crashes with ImportError after "Update complete!" (stale sys.modules in _kill_stale_dashboard_processes)
anthropics/claude-agent-sdk-python	#1220	Claude Code SDK permission system is a complete mess
TencentCloud/TencentDB-Agent-Memory	#1025	Proposal: proactive L1 memory maintenance — LLM-driven recall with source-message evidence to retire stale memories
cohere-ai/cohere-python	#796	Batched embeddings drop types absent from the first response
anthropics/claude-code	#86915	Scheduled task session hangs forever on unattended permission prompt, then per_task_limit silently starves all future runs
google/adk-python	#6729	fix(sessions): make SqliteSessionService state merges use dict.update() semantics
1jehuang/jcode	#960	Memory retrieval gaps: near-literal search misses, stale memories stay active, Chinese embeddings degrade recall
desimpson/syncer	#158	Bug: Firefox profile auto-detect fails on Windows because process.env is replaced at build time
UKGovernmentBEIS/inspect_ai	#4883	Record Task(approval=...) policies in EvalSpec (#4881)
anthropics/claude-code	#86724	Nested-Agent orchestrator (e.g. built-in /code-review) reports status: completed after yielding to spawned children, before its assigned work is done
zilliztech/memsearch	#676	Claude Code SessionStart hook still hits timeout: 10 after #645 — --version is a second full CLI boot and the PyPI check blocks (~3.4s of ~4.9s is avoidable)
anthropics/anthropic-sdk-python	#1830	fix(streaming): accumulate compaction delta content
UKGovernmentBEIS/inspect_ai	#4881	Tool Approval is not properly recorded in EvalSpec
google/adk-go	#1333	[Feature]: model/openaimodel: support image/file input (genai InlineData/FileData parts)
openai/codex	#38505	Codex App leaves thousands of zombie processes and eventually causes spawn EAGAIN
anthropics/claude-code	#86596	"Another live session" duplicate-name check fires against a session's own current name, seconds after a manual rename
51hcie/ai-project-continuity	#20	Conflict handling when multiple AI agents concurrently update tasks.md in the same session
get-bb/bb	#1552	Pool multiple Claude accounts: limit-aware load balancing across provider connections
YoanWai/agent-manager	#292	TestRefreshNotifiesWaitingTransitionOnce times out on CI
anthropics/claude-code	#86391	[BUG] Cowork scheduled tasks: WebFetch permission gate (PROVENANCE_REQUIRED) blocks unattended runs on parallel calls
openai/codex	#38350	[Bug] Recurring scheduled tasks disable themselves after successful runs without user authorization
openai/codex	#38347	Windows elevated sandbox fails on session-scoped mapped-drive cwd with error 267
google/adk-python	#6710	fix(evaluation): record NOT_EVALUATED instead of dropping invocations with zero auto-rater samples
langfuse/langfuse	#16056	bug(annotation-queues): re-adding objects creates duplicate queue items — skipDuplicates has no unique constraint to act on
mem0ai/mem0	#6935	fix(milvus): convert COSINE distance to similarity in _parse_output
mem0ai/mem0	#6933	bug(sdk-python): MilvusVectorStore returns raw distance as score for COSINE metric — ranking inverted, best match dropped by default threshold
punkpeye/fastmcp	#326	fix(edge): persist SSE responses that arrive after a client disconnect
bmad-code-org/BMAD-METHOD	#2725	[6.11.0] sprint_plan.py / sprint_status.py: BOM'd epic files silently dropped; non-scalar YAML values crash validate; CRLF files rewritten wholesale
PrismorSec/prismor	#272	Runaway agent loop detection and circuit breakers
anthropics/claude-code	#86115	Desktop app: paused (disabled) scheduled tasks disappear from the Routines list
anthropics/claude-agent-sdk-python	#1202	Keep session metadata when the first record exceeds the lite read window
planetf1/otelite	#113	feat: context-composition view — decompose cached prefix into system prompt / skills / tools / conversation
anthropics/claude-code	#85975	[Bug] Auto-update reports success with non-functional stub binary after postinstall link failure
punkpeye/fastmcp	#322	fix(edge): release cancelled GET SSE streams
ruvnet/ruflo	#2968	3.37.0 regression: memory store reports success but persists nothing — skipped better-sqlite3 postinstall silently falls back to sql.js, which rejects wal_checkpoint (re #2867, #2219)
anthropics/claude-code	#85873	[BUG] Scheduled tasks created inside an SSH remote session never run: cwd is validated against the local filesystem and no host binding is persisted
agentsmd/agents.md	#228	AGENTS.md is being used as a policy file, but two of its own rules make that unsafe
anthropics/claude-code	#85820	Permission requests in unattended scheduled-task sessions hang for days with no timeout, wedging the task queue
anthropics/claude-code	#85804	Desktop 앱 업데이트 후 계정 로컬 상태(세션 인덱스 + 예약 루틴 + Chrome 확장 페어링) 전부 리셋됨
Altinn/kihub	#117	Specify kill-switch protocol: scope, testing, and sub-agent/multi-agent coverage
rjmurillo/ai-agents	#4880	chore(context): measure and slim path-local AGENTS.md layers
anthropics/claude-code	#85663	[BUG] Windows: All install methods (npm/ps1/cmd/winget) fail with "defines.json" Syntax Error referencing C:\Program Files\nodejs
anthropics/claude-code	#85617	Model asserted third-party delivery worked from sender-side evidence; false conclusion persisted through memory across sessions (11-day production outage)
anthropics/claude-code	#85592	CLAUDE_CODE_SUBAGENT_MODEL silently discards explicit per-call subagent model since v2.1.223; documented warning never fires; subagent metadata records requested (not effective) model
anthropics/claude-code	#85565	[BUG] Desktop app update silently wiped the internal scheduled-tasks registry (scheduledTasks: []) — all scheduled tasks died at once, with zero user notification
NishikawaButterfly/quant-risk-engine	#51	Catch SystemExit in the CLI test harness
anthropics/anthropic-sdk-python	#1820	Fix streaming accumulator crash when message_start omits usage
anthropics/claude-code	#85430	[BUG] PreToolUse hooks silently do not fire for one working directory while permissions.deny from the same settings.json does
anthropics/claude-code	#85422	[FEATURE] Token-burn circuit breaker: runtime-enforced spend caps with per-source attribution (hooks, plugins, subagents), not just warnings
NousResearch/hermes-agent	#82801	Built-in personality system overrides SOUL.md — no composition, no precedence, no visible indicator
anthropics/anthropic-sdk-python	#1815	fix(streaming): initialize usage when message_start omits it
UnityInFlow/agent-learning-lab	#15	Phase 5B — Verification loops, bounded self-healing, and completion
anthropics/claude-code	#85199	[BUG]Claude Desktop repeatedly crashes and requires “Advanced Options → Repair” on Windows
dipakkrishnan/windup	#7	Clean up plist and runner on failed launchctl bootstrap
mem0ai/mem0	#6859	fix(vector_stores/azure_ai_search): raise on failed indexing results
anthropics/claude-cookbooks	#803	fix(managed-agents): bind approval to exact merge
browserbase/stagehand	#2527	fix(locator): serialize click input events
modelcontextprotocol/python-sdk	#3221	fix(server): reject concurrent duplicate JSON-RPC request ids
paperclipai/paperclip	#10412	fix(heartbeat): make agent wakeup idempotency keys actually deduplicate
openai/openai-agents-python	#3998	fix: defer session save until after output guardrails
anthropics/claude-code	#81672	fix(hookify): make package import independent of the install directory name
openai/openai-agents-python	#3984	fix(memory): prevent close() race and enforce closed state in AsyncSQLiteSession
anthropics/claude-agent-sdk-python	#1109	fix: surface prompt stream errors instead of hanging query() forever

---

Generated by github_evidence_snapshot.py on 2026-10-07T22:37:30Z. Regenerated weekly; the dated full archive lives outside GitHub.

---

🧩 One piece of a working system

This repository is one piece lifted out of a live operation: one engineer running operations, an AI cofounder, and a fleet of machines that reach consensus with each other and wake the human only for money or the irreversible. It was extracted after it survived production, not written as a demo — and it runs on its own: nothing here phones home to the rest.

See how the whole thing fits together → SYSTEM.md

Want your machine in the fleet? → Join the fleet (15 minutes, one link, no account with us)

AI contributors

This project is built by a human + AI team, and the git log says so: Claude writes most of the code, Codex and Grok review it, Gemini feeds the research. Each is credited on a commit only if its output changed that commit's content — no decorative credits. Lab-wide policy, one source for every repo: AI-CONTRIBUTORS.md.