public-skills
AGENTS.md1 starsChanged 3 months agoMITView on GitHub
Skills in this repository should follow the Agent Skills spec so they work across Claude, Codex, OpenCode, and other compatible clients. Maintainer-approved user-only skills are one explicit exception. They may use disable-model-invocation: true in SKILL.md for Claude Code and Pi, plus a minimal agents/openai.yaml with policy.allow_implicit_invocation: false for Codex. Their descriptions must also forbid automatic use as a fallback for clients that ignore these extensions. Source of truth: Before creating or editing a skill here, fetch the current docs from…
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 repository13 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- ask-claudeSkill
- battery-statsSkill
- coding-audit-assumptionsSkill
- coding-consult-designSkill
- coding-explain-problemSkill
- handoffSkill
- ntd-tasksSkill
- one-by-oneSkill
- record-macos-windowSkill
- show-meSkill
- tallySkill
- video-analysisSkill
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
- Distinctive: line 11
Before creating or editing a skill here, fetch the current docs from that index and read them.
- Distinctive: line 16
do not copy spec details into this file.
- Distinctive: line 17
keep skills portable and avoid machine-specific assumptions
- Very specific: line 34
keep description as short as possible while letting the agent select the right skill: purpose and essential selection boundaries only.
- Very specific: line 39
keep skill content independent of where it is installed: no workspace names, checkout paths, local/global installation claims, or personal …
- Distinctive: line 45
prefer minimal changes when updating an existing skill
- Very specific: line 46
do not create or regenerate agents/ UI metadata folders, including agents/openai.yaml, unless explicitly requested by the maintainer
≈ “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
Skill Authoring Rules
Skills in this repository should follow the Agent Skills spec so they work across Claude, Codex, OpenCode, and other compatible clients.
Maintainer-approved user-only skills are one explicit exception. They may use disable-model-invocation: true in SKILL.md for Claude Code and Pi, plus a minimal agents/openai.yaml with policy.allow_implicit_invocation: false for Codex. Their descriptions must also forbid automatic use as a fallback for clients that ignore these extensions.
Source of truth:
Before creating or editing a skill here, fetch the current docs from that index and read them.
Spec compliance
- the specification page must be read before authoring or changing a skill
- do not copy spec details into this file; use the upstream docs as the canonical reference
- keep skills portable and avoid machine-specific assumptions
Privacy (hard rule)
This repository is public. Do not commit:
- internal project, product, client, or organization names
- ticket IDs, codenames, or internal URLs
- credentials, tokens, API keys, or auth material
- absolute paths from any author's machine
- personal data or context that is not meant to be public
Audit every new or modified skill for leaks before commit. If unsure, do not commit.
Conventions
- one skill per folder, kebab-case name
- keep
descriptionas short as possible while letting the agent select the right skill: purpose and essential selection boundaries only; descriptions consume context before loading, so put workflow steps, implementation details, examples, and extended capability lists in the body or references - relative paths within a skill, never absolute
- keep skill content independent of where it is installed: no workspace names,
checkout paths, local/global installation claims, or personal setup instructions
in
SKILL.md, references, or bundled helpers; keep deployment details in the consuming environment's setup docs and resolve bundled resources relative to the skill itself - English for skill content unless the skill is explicitly about a non-English context
- prefer minimal changes when updating an existing skill
- do not create or regenerate
agents/UI metadata folders, includingagents/openai.yaml, unless explicitly requested by the maintainer - new coding-focused skills should use
coding-<verb>-<object>names; keep already-published names stable unless doing an explicit compatibility migration
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.