skills
AGENTS.md1 starsChanged 36 days agoMITView on GitHub
This repository authors and tests Codex-compatible agent skills. The canonical skill sources live in skills/*/SKILL.md with optional references/, scripts/, evals/, templates/, and memory/ subdirectories. Codex discovers installed copies through user skill locations such as ~/.codex/skills and ~/.agents/skills. The repo root .agents/ directory may be read-only in managed environments, so do not assume it can be edited. - For a skill change, read that skill's SKILL.md first. - Read referenced files only when the selected skill points to them for the…
- Deletes or force-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 32,671 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 repository17 other files it gives its agents.
- attentionSkill
- audit-debian-serverSkill
- frontend-qaSkill
- goalSkill
- implementSkill
- integration-smellzSkill
- invariant-minerSkill
- llm-providersSkill
- paretoSkill
- pr-describeSkill
- pr-smellzSkill
- red-teamSkill
- rust-evmSkill
- shrinkraySkill
- skill-labSkill
- student-counselSkill
- vfs-docsSkill
The file
We split this AGENTS.md into the rules it gives an agent
7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Distinctive: line 16
Read referenced files only when the selected skill points to them for the current task.
- Very specific: line 18
Use skills/common/ for shared guidance that affects multiple skills.
- Very specific: line 22
Use evals/evals.json when changing triggers, output contracts, safety rules, or behavior that should be tested.
- Distinctive: line 27
Do not inspect or summarize these unless the user explicitly asks about local tooling, caches, or runtime state:
- Distinctive: line 40
Never use whole-tree destructive recovery commands such as git reset --hard, git checkout ., or broad git restore ..
- Distinctive: line 44
Prefer patch-scoped edits and reversions.
- Very specific: line 45
Do not add bypass, evasion, credential-harvesting, or access-control circumvention instructions to skills.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 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 Guidance
Repository Purpose
This repository authors and tests Codex-compatible agent skills. The canonical
skill sources live in skills/*/SKILL.md with optional references/, scripts/,
evals/, templates/, and memory/ subdirectories.
Codex discovers installed copies through user skill locations such as
~/.codex/skills and ~/.agents/skills. The repo root .agents/ directory may
be read-only in managed environments, so do not assume it can be edited.
What To Read
- For a skill change, read that skill's
SKILL.mdfirst. - Read referenced files only when the selected skill points to them for the current task.
- Use
skills/common/for shared guidance that affects multiple skills. In skill text, reference it sibling-relative as../common/...so paths resolve both from the repo and from an installed skill directory (theskillCLI linkscommon/as a sibling of each installed skill). - Use
evals/evals.jsonwhen changing triggers, output contracts, safety rules, or behavior that should be tested.
What To Ignore
Do not inspect or summarize these unless the user explicitly asks about local tooling, caches, or runtime state:
.venv/.pytest_cache/**/__pycache__/**/*.pyc.claude/settings.local.jsonskills/**/.claude/settings.local.json- generated reports, temporary workspaces, and cache directories
Safety
- Never use whole-tree destructive recovery commands such as
git reset --hard,git checkout ., or broadgit restore .. - Preserve unrelated worktree changes. This repo is often dirty during skill development.
- Prefer patch-scoped edits and reversions.
- Do not add bypass, evasion, credential-harvesting, or access-control circumvention instructions to skills.
Verification
For any skill edit, run:
python3 scripts/check_sources.py
python3 scripts/validate_skills.pycheck_sources.py parses every Python, shell, and eval JSON file without
creating __pycache__/; validate_skills.py checks the skill contracts.
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.

