maintain-agent-metadata
Skill71 starsChanged 36 days agoMITView on GitHub
Audit, extend, or reorganize this repository's Codex guidance and skills. Use for AGENTS.md, repository skill changes, metadata placement decisions, portability checks, or instruction-context reduction.
- Reads credentials
In this repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- add-agentSkill
- analyze-tickerSkill
- diagnose-runSkill
- run-pipelineSkill
- run-testsSkill
- add-agentSkill
- analyze-tickerSkill
- diagnose-runSkill
- run-testsSkill
- validation-batchSkill
The file
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 7 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Keep only unconditional repository rules in AGENTS.md.
- Very specific: line 14
Put a bounded procedure with a recognizable trigger in one focused .agents/skills/<name>/SKILL.md.
- Very specific: line 16
Put durable architecture, rationale, or incident history in docs/ without citing agent metadata.
- Distinctive: line 20
Keep credentials, personal data, local setup, machine layout, and private notes out of every tracked destination.
- Distinctive: line 35
Keep other coding tools' metadata out of repository skills.
- Distinctive: line 37
Keep skill names short and action-oriented.
- Distinctive: line 39
Prefer instructions over scripts unless deterministic repetition justifies code.
- Distinctive: line 41
Keep root guidance below 12 KiB and avoid duplicating detail already owned by code, tests, or public documentation.
- Distinctive: line 53
When auditing ignore behavior, use git check-ignore -q AGENTS.md and require a nonzero result.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill 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
Maintain public Codex metadata
This repository's Codex metadata is public and must work in a clean clone without any user-level extension. It may not require another skill to exist.
Place material by scope
- Keep only unconditional repository rules in
AGENTS.md. - Put a bounded procedure with a recognizable trigger in one focused
.agents/skills/<name>/SKILL.md. - Put durable architecture, rationale, or incident history in
docs/without citing agent metadata. - Move cross-repository preferences or workflows to the user-level layer, but never reference that destination from tracked metadata.
- Keep credentials, personal data, local setup, machine layout, and private notes out of every tracked destination.
Audit every change
- Resolve each referenced repository path. It must be tracked, supplied by a tracked template, or clearly described as output from normal setup or execution.
- Reject ignored inputs, missing files, symlinks to outside content, absolute paths, home-relative paths, user-level skills, and local command wrappers.
.env.exampleis the only environment file whose contents may be read or quoted. Permit.envas a filename, but reject every other private.env*filename and all private environment-file contents, assignments, values, endpoints, identifiers, and holdings. Codex and metadata checks must not inspect or ingest those files; application runtime use is outside this metadata exception and must not expose values.- Keep other coding tools' metadata out of repository skills. A narrow no-touch boundary belongs in root guidance; no procedure may import or rely on that layer.
- Keep skill names short and action-oriented. The frontmatter name must equal the directory name; the description must state the actual trigger clearly.
- Prefer instructions over scripts unless deterministic repetition justifies code. Any supporting file must be tracked inside the same skill or elsewhere in the repo.
- Keep root guidance below 12 KiB and avoid duplicating detail already owned by code, tests, or public documentation.
Run:
make docs-guards
make agent-metadata-guards
poetry run pytest tests/scripts/test_agent_metadata_guards.py -v
git diff --checkWhen auditing ignore behavior, use git check-ignore -q AGENTS.md and require a
nonzero result. Do not infer ignored status from git check-ignore -v: that diagnostic
form also reports a matching negation rule.
Finish with a manual scan of the complete public metadata surface for secret-like strings, personal or holdings information, machine-specific paths, ignored-input dependencies, and references that would be meaningless in a clean clone. Keep the guard exemption count at zero.
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.

