skills
AGENTS.md12 starsChanged yesterdayMITView on GitHub
Agent Skills, installed one at a time with npx skills add or install.sh, so every skill must be self-contained. develop is the default and nightly branch, main is stable. If AGENTS.local.md exists beside this file, read it too. It holds local paths and preferences and never relaxes the rules here. Pre-push set mirroring .github/workflows/ci.yml and security.yml. Tools: shellcheck, markdownlint-cli2 (pinned in .github/tools/markdownlint/), lychee, actionlint, zizmor, gitleaks, semgrep. The last two regenerate files; commit their output. check-freshness-dates.sh and receipt age are…
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 repository40 other files it gives its agents.
- ansibleSkill
- anti-ai-proseSkill
- arch-linuxSkill
- backend-apiSkill
- ci-cdSkill
- code-reviewSkill
- code-simplificationSkill
- databasesSkill
- debian-ubuntuSkill
- debug-triageSkill
- dev-cycleSkill
- dockerSkill
- gitSkill
- i18n-localizationSkill
- kubernetes-healthSkill
- kubernetesSkill
- llm-app-developmentSkill
- mcpSkill
- networkingSkill
- nixosSkill
- observabilitySkill
- opnsense-pfsenseSkill
- plan-reviewSkill
- privilege-escalationSkill
- prompt-generatorSkill
- repo-auditSkill
- rhel-fedoraSkill
- roadmapSkill
- security-auditSkill
- session-handoffSkill
- shell-scriptingSkill
- skill-creatorSkill
- skill-refinerSkill
- synology-dsmSkill
- terraformSkill
- testingSkill
- ui-buildSkill
- update-docsSkill
- virtualizationSkill
- vulnerability-researchSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 6
If AGENTS.local.md exists beside this file, read it too.
- Distinctive: line 49
No AI attribution in commits, PRs, tags, or files.
- Distinctive: line 50
No private hosts, domains, IPs, machine paths, or work context in tracked files.
- Distinctive: line 54
Keep protected overlays and private markers out of commits.
≈ “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
skills
Agent Skills, installed one at a time with npx skills add or install.sh, so every skill must be
self-contained. develop is the default and nightly branch, main is stable.
If AGENTS.local.md exists beside this file, read it too. It holds local paths and preferences and never
relaxes the rules here.
Commands
Pre-push set mirroring .github/workflows/ci.yml and security.yml. Tools: shellcheck,
markdownlint-cli2 (pinned in .github/tools/markdownlint/), lychee, actionlint, zizmor, gitleaks, semgrep.
./scripts/lint-skills.sh
./scripts/validate-spec.sh
for f in scripts/check-*.sh; do "$f" || echo "FAIL $f"; done
python3 .github/stack/stack-check.py
for f in scripts/test-*.sh; do "$f" || echo "FAIL $f"; done
python3 scripts/test-migrations.py
shellcheck install.sh scripts/*.sh skills/repo-audit/references/detect.sh
markdownlint-cli2
lychee --config lychee.toml README.md SECURITY.md CONTRIBUTING.md CHANGELOG.md INSTALL.md MIGRATION.md AGENTS.md 'skills/**/*.md' 'docs/**/*.md'
actionlint
zizmor .
gitleaks git --redact .
semgrep scan --metrics off --error --config p/ci --config p/github-actions install.sh scripts .github/workflows .github/stack skills/repo-audit/references/detect.sh
./scripts/gen-contract-refs.sh
python3 scripts/gen-ref-toc.pyThe last two regenerate files; commit their output. check-freshness-dates.sh and receipt age are
advisory. pre-commit install --hook-type pre-commit --hook-type pre-push wires the local hooks,
which run a subset of this block (the list of what they skip heads .pre-commit-config.yaml).
Layout
skills/<name>/: one installable skill. Never reference paths outside it at runtime.skills/_shared/: build-time sources only. Each skill'sreferences/output-contract.mdandagent-hygiene.mdare generated from them; never hand-edit those copies.scripts/: gates (check-*.sh,lint-skills.sh,validate-spec.sh) and their tests (test-*)..refiner-runs.json,.refiner-ledger.md: skill-refiner history at the root by contract; append only, except that private paths are removed..github/stack/: vendored from agents; do not edit.- Details: CONTRIBUTING.md, INSTALL.md, MIGRATION.md.
Rules
- Issue first, branch from
develop, PR intodevelop, conventional commits,Closes #NN. Never commit tomainordevelop. - No AI attribution in commits, PRs, tags, or files.
- No private hosts, domains, IPs, machine paths, or work context in tracked files.
SKILL.mdstays at 500 lines or fewer; every reference is linked from it.- A new active skill needs repo-audit coverage (see CONTRIBUTING.md); renames and removals need
a
migrations.jsonentry and aMIGRATION.mdchange. - Keep protected overlays and private markers out of commits.
For agents
./install.sh --json(installs,--list,--check,--detect,--doctor,--migrate,--update) prints one JSON object; keys and exit codes are in INSTALL.md.- There is no single all-checks command; the block above is the full set and CI is the source of truth.
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.

