Add commands

Document ante and juror from HN discovery; block nixpkgs ante collision.

Author: SimonSchubert

assets/commands/ante.md

@@ -0,0 +1,146 @@
+# TAGLINE
+
+Self-contained terminal coding agent with multi-provider and offline support
+
+# TLDR
+
+**Launch interactive terminal UI**
+
+```ante```
+
+**Run a one-shot headless task**
+
+```ante -p "[add error handling to src/main.rs]"```
+
+**Pipe a diff for review**
+
+```git diff | ante -p "[review this for security issues]"```
+
+**Use a specific provider and model**
+
+```ante --provider [openai] --model [gpt-5.5] -p "[refactor the database module]"```
+
+**Run fully offline with a local GGUF model**
+
+```ante --offline-model [~/.ante/models/model.gguf] -p "[add tests]"```
+
+**Resume a saved session**
+
+```ante --resume [ses_...] -p "[now add tests]"```
+
+**Update to the latest release**
+
+```ante update```
+
+# SYNOPSIS
+
+**ante** [_options_] [**-p** _prompt_]
+
+**ante** **serve**
+
+**ante** **gateway**
+
+**ante** **update** [_options_]
+
+# PARAMETERS
+
+**-p**, **--prompt** _PROMPT_
+> Run in headless mode with the given prompt and exit when done
+
+**-m**, **--model** _MODEL_
+> Override the model name
+
+**--provider** _PROVIDER_
+> Override the API provider (for example anthropic, openai, gemini, xai, openrouter, local)
+
+**--profile** _NAME_
+> Use **~/.ante/**_NAME_**.settings.json** instead of the default settings; **ANTE_PROFILE** is equivalent
+
+**--effort** _LEVEL_
+> Override model effort: min, low, medium, high, xhigh, or max
+
+**--yolo**
+> Skip all interactive tool approval prompts
+
+**--output-format** _FORMAT_
+> Headless output format: json, human, or minimal (default: minimal)
+
+**--offline-model** _PATH_
+> Path to a local GGUF model file; boots built-in local inference and uses the local provider
+
+**-r**, **--resume** _SESSION_ID_
+> Resume a previously saved session by ID
+
+**--no-session-save**
+> Do not persist a transcript or resumable snapshot
+
+**--check**
+> After the main task, run a verification pass on the work
+
+**--system-prompt** _PROMPT_
+> Replace the default system prompt entirely
+
+**--system-prompt-file** _PATH_
+> Read a replacement system prompt from a UTF-8 file
+
+**--append-system-prompt** _TEXT_
+> Append text to the system prompt
+
+**--no-skills**
+> Skip skill discovery; no skills are advertised or invocable
+
+**--tools** _TOOLS_...
+> Replace the default tool set with exactly these tools
+
+**--include-tools** _TOOLS_...
+> Add tools on top of the default or **--tools** set
+
+**--exclude-tools** _TOOLS_...
+> Remove tools after **--tools** and **--include-tools**
+
+# COMMANDS
+
+**serve**
+> Run as a long-lived daemon over a structured JSONL protocol for editor plugins and integrations
+
+**gateway**
+> Run Ante as a Slack or Discord bot
+
+**update**
+> Install the latest release; supports **--channel** and **--version**
+
+# DESCRIPTION
+
+**ante** is a self-contained coding agent from Antigma Labs that runs in the terminal. It is distributed as a single Rust binary (~15MB) with no runtime language dependencies. Day-to-day use is an interactive TUI (**ante** with no arguments); headless one-shots use **-p** / **--prompt**.
+
+It works with many cloud providers (Anthropic, OpenAI, Gemini, Grok/xAI, Open Router, and OpenAI-compatible endpoints) via API keys or OAuth, and can also run fully offline against a local GGUF model through a built-in inference engine. Configuration lives under **~/.ante/** (settings, catalog, and optional named profiles).
+
+Subcommands cover programmatic embedding (**ante serve**), chat-platform bots (**ante gateway**), and self-updates (**ante update**). Features include multi-agent orchestration, skills, MCP servers, and persistent memory across sessions.
+
+# CAVEATS
+
+Ante is in alpha/preview: expect breaking changes and incomplete functionality. Official support is macOS and Linux (Windows users are directed to WSL). Headless mode implies auto-approval of tool calls (yolo). Cloud providers need credentials (for example **ANTHROPIC_API_KEY**, **OPENAI_API_KEY**); offline mode needs a local GGUF. The prebuilt binary is free to use during the alpha under separate binary terms; open-source parts of the repository are Apache-2.0. Install via the official script (**curl -fsSL https://ante.run/install.sh | bash**); distro packages named **ante** are typically a different project (the Ante programming language).
+
+# CONFIGURATION
+
+Default settings: **~/.ante/settings.json**
+
+Named profiles: **~/.ante/**_name_**.settings.json** (select with **--profile** or **ANTE_PROFILE**)
+
+Provider/model catalog overrides: **~/.ante/catalog.json**
+
+Binary install location defaults to **~/.ante/bin** (override with **ANTE_INSTALL_DIR** during install)
+
+# SEE ALSO
+
+[claude](/man/claude)(1), [codex](/man/codex)(1), [opencode](/man/opencode)(1), [aider](/man/aider)(1), [goose](/man/goose)(1)
+
+# RESOURCES
+
+```[Source code](https://github.com/AntigmaLabs/ante)```
+
+```[Homepage](https://antigma.ai)```
+
+```[Documentation](https://docs.antigma.ai)```
+
+<!-- verified: 2026-08-10 -->

assets/commands/index.txt

@@ -215,6 +215,7 @@ ansible-vault.md
 ansible.md
 ansiweather.md
 ant.md
+ante.md
 anthropic.md
 antibody.md
 antigen.md
@@ -3828,6 +3829,7 @@ jupyter.md
 jupyterlab.md
 jupyterlite.md
 jupytext.md
+juror.md
 just.1.md
 just.js.md
 just.md

assets/commands/juror.md

@@ -0,0 +1,105 @@
+# TAGLINE
+
+Multi-model AI pull request review CLI with cost receipts
+
+# TLDR
+
+**Review the current branch against a base**
+
+```juror review --base [main]```
+
+**Review a pull request and print findings**
+
+```juror review --pr [1234] --repo [owner/name]```
+
+**Review a PR and post the sticky comment**
+
+```juror review --pr [1234] --repo [owner/name] --post```
+
+**Use a jury preset**
+
+```juror review --preset [balanced] --base [main]```
+
+**Run a shadow benchmark from a corpus file**
+
+```juror benchmark --file [benchmarks/case.json]```
+
+# SYNOPSIS
+
+**juror** **review** [_options_]
+
+**juror** **benchmark** [**--file** _path_]
+
+# PARAMETERS
+
+**review**
+> Run a multi-model review of a local branch or a GitHub pull request
+
+**--base** _REF_
+> Base ref to diff against for a local working-branch review
+
+**--pr** _NUMBER_
+> Pull request number to review
+
+**--repo** _OWNER/NAME_
+> GitHub repository for **--pr** reviews
+
+**--post**
+> Post (or update) the sticky review comment on the pull request
+
+**--preset**, **--mode** _NAME_
+> Jury preset: fast (default), balanced, high, or ultra
+
+**--models** _A,B,..._
+> Narrow the selected preset (or custom jury) to these model ids for one run
+
+**--config** _PATH_
+> Path to a Juror config file (default: **.juror.yml** at the repo root)
+
+**--cost-target** _USD_
+> Per-PR planning cost target in USD
+
+**--repo-dir** _PATH_
+> Local repository directory (used by CI and advanced invocations)
+
+**--json** _PATH_
+> Write structured review output as JSON to _PATH_
+
+**--dry-run**
+> Do not treat the run as a live publish (used with Action dry runs)
+
+**benchmark**
+> Evaluate reviewer quality against an adjudicated corpus file
+
+**--file** _PATH_
+> Benchmark corpus JSON file
+
+# DESCRIPTION
+
+**juror** (npm package **juror-ai**) is a multi-model code review tool that runs several frontier models in parallel through their native agent harnesses (Claude Code, Codex, OpenCode, Grok Build, Kimi Code, and generic OpenAI-compatible endpoints), then merges and deduplicates findings into one report with a merge-confidence score and a per-model cost receipt.
+
+It is designed for GitHub pull requests and local branch diffs. Typical install is **npm i -g juror-ai** (binary names **juror** and **juror-ai**). The same code path powers the **juror-ai/juror** GitHub Action, which posts a sticky summary comment and batched inline review comments on PRs.
+
+Findings are anchored to the diff, blocked by file/line window, collapsed by exact and similarity/referee stages, and optionally filtered for higher precision via **publish_mode: consensus**. Provider keys are isolated per harness; models never receive **GITHUB_TOKEN**.
+
+# CAVEATS
+
+Requires Node.js 20+ and at least one LLM provider API key (**JUROR_OPENAI_API_KEY**, **JUROR_ANTHROPIC_API_KEY**, **JUROR_XAI_API_KEY**, **JUROR_FIREWORKS_API_KEY**, or unprefixed vendor fallbacks). Actual review quality and cost depend on installed harness CLIs and available keys. Cost for some harnesses is estimated from tokens × list price rather than provider-reported USD. Fork PRs in GitHub Actions are skipped by design when secrets are withheld. Not an autofix bot or linter.
+
+# CONFIGURATION
+
+Repo-root **.juror.yml** (optional). Keys include **preset**, **models**, **consensus**, **review** (publish_mode, severity_floor, paths_ignore), **budget**, and **output**. An explicit **models:** list replaces the preset entirely.
+
+Provider keys in environment or a local **.env** (loaded automatically; untracked files are not exposed to model checkouts).
+
+# SEE ALSO
+
+[gh](/man/gh)(1), [claude](/man/claude)(1), [codex](/man/codex)(1), [opencode](/man/opencode)(1), [grok](/man/grok)(1)
+
+# RESOURCES
+
+```[Source code](https://github.com/Juror-AI/juror)```
+
+```[Documentation](https://github.com/Juror-AI/juror#readme)```
+
+<!-- verified: 2026-08-10 -->

scripts/fetch-package-installs.py

@@ -126,6 +126,8 @@
         "aur": "mcpp-bin",
         "brew": "mcpp-community/mcpp/mcpp-m",
     },
+    # ante: nixpkgs "ante" is jfecher/ante (programming language), not AntigmaLabs coding agent
+    "ante": {},
 }
 
 # Prefer these path prefixes when mapping binaries (order = priority).