agents-skills
AGENTS.md2 starsChanged 13 days agoMITView on GitHub
This repository builds a commercial catalog of curated Upstream Skills and versioned Solution Skills. Codex and Claude Code are the only supported hosts. Do not add targets, manifests, or setup instructions for other agents. - catalog.yaml indexes public upstream records, bundled skills, discovery profiles, and released solutions. - solutions/ contains Solution Skill manifests and locked component digests. - skills/ contains bundled skill packages. Their SKILL.md files link detailed references and scripts. - .codex/agents/ and .claude/agents/ contain host-specific custom agent definitions.…
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 31,439 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 repository10 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- design-taste-frontendSkill
- eraser-diagramsSkill
- codebase-onboarding-c4Skill
- copywritingSkill
- frontend-designSkill
- html-docs-standardSkill
- html-slidesSkill
- presentation-brandSkill
- problem-presentationSkill
The file
We split this AGENTS.md into the rules it gives an agent
7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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 29
Treat an upstream commit and its skill digest as immutable.
- Very specific: line 35
Keep each host's installation path separate: Codex uses .agents/skills/ and Claude Code uses .claude/skills/.
- Distinctive: line 38
Do not pin a model in skill metadata or agent definitions without a specific product reason.
- Very specific: line 40
Keep skill entrypoints concise and link to detailed references.
- Very specific: line 43
Do not represent the html-slides rewrite's original upstream derivation commit as known.
- Very specific: line 51
Run npm run format:check, npm run lint, npm run typecheck, npm test, npm run build, and git diff --check.
- Distinctive: line 60
Start with README.md, then the catalog and technical guide.
≈ “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
Repository instructions
This repository builds a commercial catalog of curated Upstream Skills and versioned Solution Skills. Codex and Claude Code are the only supported hosts. Do not add targets, manifests, or setup instructions for other agents.
Source of truth
catalog.yamlindexes public upstream records, bundled skills, discovery profiles, and released solutions.solutions/contains Solution Skill manifests and locked component digests.skills/contains bundled skill packages. TheirSKILL.mdfiles link detailed references and scripts..codex/agents/and.claude/agents/contain host-specific custom agent definitions..agents/skills/contains local authoring skills; Claude Code links to them from.claude/skills/.src/is the TypeScript CLI for source staging, validation, release builds, installs, and upgrades.docs/product-model.mddefines the two product types.docs/agent-catalog.mdrecords current release status.docs/guides/technical-v1.mddocuments the publisher and customer workflow.DESIGN.mddefines the shared NKSkills visual system for owned pages and materials. Customer-generated artifacts use their own brand.
Working agreements
- Treat an upstream commit and its skill digest as immutable. Review new
upstream content, attribution, license, and compatibility before changing a
pin. Never let
check-upstreamsupdate a release automatically. - A Solution Skill has its own version, outcome, entrypoint, quality gates, and complete component lock. Do not call an experimental catalog profile a commercial release.
- Keep each host's installation path separate: Codex uses
.agents/skills/and Claude Code uses.claude/skills/. Verify current CLI behavior against official host documentation before changing these paths or commands. - Do not pin a model in skill metadata or agent definitions without a specific product reason. Let the host select its configured model.
- Keep skill entrypoints concise and link to detailed references. Preserve the
read-only evidence boundary in
codebase-onboarding-c4; durable writes and external access remain governed by its approval and quality contracts. - Do not represent the
html-slidesrewrite's original upstream derivation commit as known. The current upstream record is a reviewed baseline only. - For NKSkills-owned UI, read
DESIGN.mdbefore editing. Keep its roles synchronized withsite/src/styles/tokens.cssand run the site design contract tests. Customer artifacts keep their own brand profile.
Verification
Run npm run format:check, npm run lint, npm run typecheck, npm test,
npm run build, and git diff --check. When a skill changes, validate its
package and rebuild any release whose locked digest changes. For a release
change, build the versioned archive and install the extracted archive into a
fresh project for both hosts. Check links, license notices, component digests,
and the project installation lock.
For changes to the Eraser render script, run
python3 -m unittest discover -s tests -v.
Start with README.md, then the catalog and technical guide.
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.

