trac3r00trac3r00/agent-skills/AGENTS.md

agent-skills

AGENTS.md0 starsChanged 3 months agoMITView on GitHub

Generated: 2026-08-26 Commit: 184b217 Branch: main

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 repository69 other files it gives its agents.
- skills/AGENTS.mdAGENTS.md
- skills/ultimate-browsing/engine/AGENTS.mdAGENTS.md
- 3d-designSkill
- algorithmic-artSkill
- api-testerSkill
- apple-suiteSkill
- appshotSkill
- astra-initSkill
- blindspotSkill
- changelog-genSkill
- claim-auditSkill
- code-reviewSkill
- comment-checkerSkill
- context-budgetSkill
- cron-auditSkill
- css-pro-tipsSkill
- designSkill
- diff-reviewSkill
- doc-readerSkill
- domain-modelingSkill
- env-gateSkill
- ext-scaffoldSkill
- gate-graphSkill
- gh-address-commentsSkill
- gh-fix-ciSkill
- git-masterSkill
- graph-toolsSkill
- grillingSkill
- herdrSkill
- image-genSkill
- json-diffSkill
- linus-levelSkill
- log-analyzerSkill
- log-deviationSkill
- lsp-setupSkill
- merge-quizSkill
- nbj-write-clearlySkill
- net-probeSkill
- open-loopsSkill
- portfolio-auditSkill
- pro-reviewSkill
- proxmox-veSkill
- remove-ai-slopsSkill
- repo-auditSkill
- resume-auditSkill
- sec-headersSkill
- secret-gateSkill
- seo-auditSkill
- session-finderSkill
- session-handoffSkill
- session-rulesSkill
- skill-auditSkill
- skill-creatorSkill
- skill-decaySkill
- skill-optimizerSkill
- skill-pickerSkill
- skill-syncSkill
- social-researchSkill
- subscription-auditSkill
- sys-healthSkill
- The first 60 shown.

The file

We split this AGENTS.md into the rules it gives an agent

3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
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

Very specific:
No network/credentials/LLM calls inside guard CLIs — offline and deterministic, always.
line 46
Very specific:
Never overwrite user files in sync-type tools (skill-sync skip-exists.
line 47
Very specific:
Don't hand-edit one manifest: .claude-plugin/marketplace.json, .claude-plugin/plugin.json, .codex-plugin/plugin.json move together.
line 49

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

AGENTS.md

PROJECT KNOWLEDGE BASE

Generated: 2026-08-26 Commit: 184b217 Branch: main

OVERVIEW

61 agent skills (SKILL.md standard) in five families; 35 are script-backed stdlib-only Python CLIs with exit-code gates, 22 are instruction-only. Installable as a Claude Code / Codex plugin or used from a clone by any SKILL.md-reading agent.

STRUCTURE

agent-skills/
├── .claude-plugin/   # marketplace.json (catalog) + plugin.json (metadata), version-locked together
├── .codex-plugin/    # Codex manifest; skills path ./skills/, no version field
├── .agents/plugins/  # generic marketplace manifest
├── skills/<name>/    # SKILL.md [+ scripts/*.py] [+ references/] [+ LICENSE.txt]
└── tests/test_skills.py  # single test module for ALL skill CLIs + repo-wide frontmatter validation

WHERE TO LOOK

Task	Location	Notes
Add a skill	`skills/<name>/SKILL.md`	frontmatter contract in skills/AGENTS.md; validated by tests
Add CLI tests	`tests/test_skills.py`	script const near `CB`..`SH`, use `run()` helper
Bump release	`.claude-plugin/*.json`, `.codex-plugin/plugin.json`	keep versions + descriptions in sync across all three
Skill families / inventory	`README.md` tables	update when adding a skill; counts are asserted nowhere — keep honest
CI	`.github/workflows/tests.yml`	pytest matrix 3.9/3.11/3.12 + 2 smoke tests

CODE MAP

Symbol	Type	Location	Role
`run(script,*args,stdin)`	helper	tests/test_skills.py:19	subprocess runner all CLI tests share
`test_every_skill_has_valid_frontmatter`	test	tests/test_skills.py	repo-wide gate: name==dir, description present
`discover()`	fn	skills/skill-sync/scripts/skill_sync.py	provider-root scan, priority dedupe
`SCANNERS`	dict	skills/session-handoff/scripts/session_handoff.py	per-client session store readers
`PROVIDER_ROOTS`	const	skills/skill-sync/scripts/skill_sync.py	priority-ordered skill dirs per provider

CONVENTIONS

CLI contract (all 35 script-backed skills): argparse, positional input or - stdin, --json, gates are strict > comparisons; rc 0 ok / 1 gate breach / 2 usage-or-input error.
Stdlib-only; sole exception: context-budget optionally uses tiktoken (chars/4 fallback).
Frontmatter: name (== dirname), description, version, license (with upstream attribution), metadata.agentskills.tags.
Ported skills keep upstream LICENSE.txt in-dir and a license: frontmatter line naming the source repo.
De-brand ported content: harness-specific tools named only as examples with plain-text fallbacks.

ANTI-PATTERNS (THIS PROJECT)

No network/credentials/LLM calls inside guard CLIs — offline and deterministic, always.
Never overwrite user files in sync-type tools (skill-sync skip-exists; session-handoff is read-only).
Tool claims stay honest: claim-audit is NOT a fact-checker; skill-decay "never used" is a review prompt, not proof; gate-graph overlap is a candidate, not semantic equivalence.
Don't hand-edit one manifest: .claude-plugin/marketplace.json, .claude-plugin/plugin.json, .codex-plugin/plugin.json move together.

COMMANDS

uvx pytest tests/ -q                 # full suite (112 tests)
python3 skills/<s>/scripts/<s>.py --help

NOTES

skills/ultimate-browsing/engine/AGENTS.md documents the vendored fetch engine — leave it upstream-shaped.
macOS: session-handoff normalizes /tmp -> /private/tmp; cwd matching is target-or-child, never parent.
CI tests Python 3.9 while README floor is 3.9 too — keep them aligned when bumping.

Download now (ZIP)

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

What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free