Claude-Code-Plugins
AGENTS.md8 starsChanged 2 months agoMITView on GitHub
Session state: Agent Handoff SessionStart injects docs/handoff/state.md; do not reread it when injected. Then use this file and docs/handoff/conventions.md. Full conventions reference: docs/handoff/conventions.md - LLM-targeted pattern library. Every convention follows the six-field schema (Applies-when / Rule / Code / Why / Sources / Related) with a Quick Reference table at the top for O(1) lookup. Do not introduce new patterns without checking conventions first. Detailed review workflows: AGENTS.reviews.md - read this only for review-related tasks (review planning, review sweeps, code/security/test/etc.…
- Commits and pushes
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository5 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- agent-handoffSkill
- agent-handoffSkill
- refresh-plugin-cacheSkill
- test-pluginSkill
The file
We split this AGENTS.md into the rules it gives an agent
11 rules. About 0 are common, 6 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 15
Treat the design specs indexed in docs/handoff/specs-plans.md (stored under docs/plans/, docs/research/, and …
- Very specific: line 16
Keep .claude-plugin/plugin.json, plugin folders, command wiring, and marketplace metadata in sync.
- Distinctive: line 27
When changing Markdown, JSON, JSONC, or YAML, run the fix pass first:
- Distinctive: line 36
Before considering work complete, run the non-mutating check:
- Distinctive: line 43
Do not claim completion if either command fails.
- Very specific: line 49
Do not edit .prettierrc.json or .markdownlint.json to bypass a check without a documented ADR exception.
- Very specific: line 57
Use the repo-local agent-handoff skill at session startup and closeout.
- Distinctive: line 76
Check formatting over exactly that scope, with Git as the corpus authority:
- Very specific: line 88
Never check or write with a bare .: it reaches undeclared languages and Git-excluded scratch.
- Distinctive: line 96
Never lint a bare recursive glob: it descends into any independent Git repository checked out below this one.
- Distinctive: line 98
Run the enabled checks before claiming completion.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Codex Instructions for Claude-Code-Plugins
Session state: Agent Handoff SessionStart injects docs/handoff/state.md; do not reread it when injected. Then use this file and docs/handoff/conventions.md.
Full conventions reference: docs/handoff/conventions.md - LLM-targeted pattern library. Every convention follows the six-field schema (Applies-when / Rule / Code / Why / Sources / Related) with a Quick Reference table at the top for O(1) lookup. Do not introduce new patterns without checking conventions first.
Detailed review workflows: AGENTS.reviews.md - read this only for review-related tasks (review planning, review sweeps, code/security/test/etc. reviews). The verbose per-review routing, defaults, and orchestrator notes live there.
Repo Purpose
Plugin authoring and release workspace for Claude Code / Codex plugins.
Key Rules
- Treat the design specs indexed in
docs/handoff/specs-plans.md(stored underdocs/plans/,docs/research/, anddocs/superpowers/{specs,plans}/) as the architectural source of truth for plugin behavior; the marketplace schema lives indocs/handoff/architecture.md. - Keep
.claude-plugin/plugin.json, plugin folders, command wiring, and marketplace metadata in sync. - Validate substantive plugin changes with the plugin test harness before wrapping up.
- Preserve documented enforcement layers, hooks, and release expectations when refactoring.
- Branch workflow: direct commit to
main. Notestingbranch — that convention was retired 2026-05-07. Plugin releases are manual: bumpplugin.jsonand the matching.claude-plugin/marketplace.jsonentry, commit, tag<name>/vX.Y.Z, push, andgh release create. See BRANCH_PROTECTION.md.
Markdown & Structured-Text Tooling
This repository follows the Markdown Tooling Standard (project-standards v3.0.0). Prettier formats the structured-text it supports (md/json/jsonc/yaml/code-workspace); markdownlint lints Markdown structure only. JS/TS source is excluded from Prettier via .prettierignore. MD060 is enabled at {style: "leading_and_trailing", aligned_delimiter: false} (Prettier-compatible; the standard's {style: "any"} conflicts with Prettier's empty-cell rendering). Both deviations recorded in docs/decisions/adr-0001-prettier-jsts-scope.md. Do not introduce a competing formatter or linter.
Fix pass
When changing Markdown, JSON, JSONC, or YAML, run the fix pass first:
npx prettier --write .
npx markdownlint-cli2 --fix "**/*.md"Check contract
Before considering work complete, run the non-mutating check:
npx prettier --check .
npx markdownlint-cli2 "**/*.md"Do not claim completion if either command fails.
Rules
- Prettier owns physical formatting. Do not fight its output or hand-format.
- markdownlint owns Markdown structure. Do not disable a rule to silence a warning — fix the Markdown.
- Do not edit
.prettierrc.jsonor.markdownlint.jsonto bypass a check without a documented ADR exception.
<!-- prettier-ignore-start --><!-- BEGIN project-standards:agent-handoff --><!-- markdownlint-disable MD025 -->Agent Handoff
Use the repo-local agent-handoff skill at session startup and closeout. Do not reread state already injected by SessionStart. Keep project knowledge inside this repository and store credential references only, never values.
<!-- markdownlint-enable MD025 --><!-- END project-standards:agent-handoff --><!-- prettier-ignore-end --><!-- prettier-ignore-start --><!-- BEGIN project-standards:markdown-tooling --><!-- markdownlint-disable MD025 -->Markdown and structured-text tooling
Prettier owns physical formatting and markdownlint owns Markdown structure. Do not add overlapping tools.
Enabled checks: format, lint.
Markdown scope: **/*.md.
Structured-config scope: **/*.json, **/*.jsonc, **/*.yml, **/*.yaml.
Lint additionally skips generated directories: .pytest_cache/**, .ruff_cache/**, .venv/**, node_modules/**.
Check formatting over exactly that scope, with Git as the corpus authority:
git ls-files -z -- ':(glob)**/*.md' ':(glob)**/*.json' ':(glob)**/*.jsonc' ':(glob)**/*.yml' ':(glob)**/*.yaml' | xargs -0 -r npx prettier --check --Without Git, bound the same scope by glob instead:
npx prettier --check --no-error-on-unmatched-pattern -- '**/*.md' '**/*.json' '**/*.jsonc' '**/*.yml' '**/*.yaml'Never check or write with a bare .: it reaches undeclared languages and Git-excluded scratch.
Lint Markdown structure over the same Git-tracked scope:
git ls-files -z -- ':(glob)**/*.md' ':(glob,exclude).pytest_cache/**' ':(glob,exclude).ruff_cache/**' ':(glob,exclude).venv/**' ':(glob,exclude)node_modules/**' | sed -z 's|^|:|' | xargs -0 -r npx markdownlint-cli2 --no-globsNever lint a bare recursive glob: it descends into any independent Git repository checked out below this one.
Run the enabled checks before claiming completion.
<!-- markdownlint-enable MD025 --><!-- END project-standards:markdown-tooling --><!-- prettier-ignore-end -->Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

