Repository navigation
chore(instruction-placement): address Claude Code v2.1.265..v2.1.296 changelog - #6780
Conversation
…changelog Row 33 (2.1.288-061, with 2.1.293-039) and the scope of #6199: path-scoped rules and nested CLAUDE.md no longer load only on Read. - render-index.sh: the rendered index preamble now says a surface loads when Claude works with a file it covers and links memory#path-specific-rules for which tool calls count, instead of listing them; pointer record (as-of 2026-10-10, recheck trigger) added beside the heredoc; header comments updated to match. - AGENTS.md: regenerated with render-index.sh write (preamble, plus the pending tests/fixtures row from #6669). - realign SKILL.md, README.md: read-only trigger wording replaced with a pointer to context/verified-mechanics.md; creation deny unchanged. - adherence-experiment.sh and adherence-results.md: trigger wording only; recorded results untouched. Refs: #6199 Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
skills/migrate/reference/verification.md said a nested AGENTS.md attaches on a Read "and nowhere else". It now says the file attaches when Claude works with a file there, and adds a source-conflict record: changelog v2.1.290 and memory#agents-md disagree on what attaches a subdirectory AGENTS.md (as of 2026-10-10, with a recheck trigger). Refs: #6199 Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…agment Refs: #6199 Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…NTS.md The rendered index preamble and the migrate verification guide said a surface or nested AGENTS.md loads when Claude "works with" a covered file, which implies Write/Edit triggers no source supports for AGENTS.md. The preamble now says a surface loads once a covered file triggers it and links memory#path-specific-rules (rules, nested CLAUDE.md) and memory#when-claude-code-reads-agents-md (AGENTS.md); the verification guide states no trigger set and points its source-conflict record at the AGENTS.md subsection. AGENTS.md regenerated with render-index.sh write; check is IN-SYNC. Refs: #6199 Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Reword the remaining "until a read happens to match" and "read again" phrasings in routing-rubric.md, verified-mechanics.md, index-drift.sh, the check skill's evals.json and an adherence-experiment.test.sh comment, without listing triggers. The rendered index preamble and its pointer record now link each surface type to its own memory section: path-specific-rules (rules), how-claude-md-files-load (nested CLAUDE.md), when-claude-code-reads-agents-md (AGENTS.md). AGENTS.md regenerated; render-index.sh check is IN-SYNC. Refs: #6199 Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Claude finished @kyle-sexton's task in 3m 21s —— View job Code review complete
This is a docs/wording-accuracy PR for the One finding posted inline on Everything else I checked — the |
|
Claude finished @kyle-sexton's task in 1m 38s —— View job Security review
No security issues found. This PR changes only documentation/comment text and generator output in the Instruction-surface deletion check: several of the changed files are standing-instruction |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 88e310e165
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
claude-security-review has reviewed this pull request through 58f123f; a later push is reviewed from there. |
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
|
claude-review has reviewed this pull request through bd294af; a later push is reviewed from there. |
…tive path in realign The changed-skills gate resolves backticked paths against the skill directory; context/verified-mechanics.md lives two levels up. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Claude finished @kyle-sexton's task in 56s —— View job Code review complete
Scope: per this lane's instructions, only the change since What changed: commit Verification:
No findings. This is a correct, narrowly-scoped one-line path fix with no remaining review surface; nothing to flag. |
|
Claude finished @kyle-sexton's task in 42s —— View job Security review
Scope: incremental. Per the lane header, this lane last reviewed The change: one line, fixing the relative path in the "Forgetting the index regeneration" gotcha bullet from No security issues found.
|
…sm with plain statement Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ict in verified-mechanics verification.md now points at that record instead of carrying its own. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Claude finished @kyle-sexton's task in 1m 10s —— View job Code review complete
Scope: per this lane's instructions, only what changed since What changed:
Verification:
No findings. This is a well-scoped follow-up that resolves the prior round's finding without introducing new inconsistencies. |
|
Claude finished @kyle-sexton's task in 1m 14s —— View job Security review complete
Scope: incremental. Per the lane header, this lane last reviewed
Every other file in this PR was already reviewed at The changes:
This closes the finding from the earlier full review on No security issues found.
|
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
1 participant
Closes #6199
Summary
The instruction-placement plugin, and the root index preamble that
render-index.shgenerates into every consuming repo's always-loaded file, said path-scoped surfaces load only when Claude reads a matching file. That has been false for path-scoped rules and nested CLAUDE.md since Claude Code 2.1.288, which also loads them on write and edit. The plugin and the preamble no longer say load-on-read only; each surface type now points at its own memory docs section for the exact triggers. The rootAGENTS.mdchanged only as generator output.Fix
plugins/instruction-placement/scripts/render-index.sh: the generated preamble says a surface loads once a file it covers triggers it and links the memory docs sections for rules, nested CLAUDE.md and AGENTS.md instead of listing triggers;hooks/index-drift.shmatches.README.md,skills/realign/SKILL.md,skills/check/evals/evals.json,context/routing-rubric.md,context/verified-mechanics.md,evals/adherence-experiment.sh,evals/adherence-experiment.test.sh,evals/adherence-results.md: the read-only trigger wording is replaced by a pointer toverified-mechanics.md; recorded adherence results are unchanged, and creation-governing content is still denied the path-scoped destination.skills/migrate/reference/verification.md: no longer says a nested AGENTS.md attaches only on a Read; records the changelog and memory docs disagreement as a source-conflict pointer.AGENTS.md: regenerated preamble only..changes/instruction-placement/.Verification
render-index.shcheck: IN-SYNC.render-index.test.sh(5 failures) andadherence-experiment.test.sh(2 failures) fail on main on Windows (symlink and drive-letter cases) and are unchanged by this PR.Related
Part of the Claude Code v2.1.265..v2.1.296 changelog apply (one PR per plugin; the upstream
ledger PR comes last). Related: #6193 (creation deny, out of scope) and #6103.
🤖 Generated with Claude Code