skills
AGENTS.md0 starsChanged 17 days agoMITView on GitHub
This repository is the public home of the skills (skills/<name>/). It is content only: no package manager, no build step, no test runner. Verify a skill with python3 scripts/check-skills.py, which fails on missing frontmatter, interface metadata, README rows, broken relative links, a CREDITS.md without its LICENSE, a description that claims a sibling's trigger, or a <!-- check:... --> tag that names no code its checker emits. Verify a figure with python3 skills/tech-diagrams/scripts/self_check.py <source> --png <render>. When you write or edit…
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 31,700 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 repository20 other files it gives its agents.
- architecture-ablationSkill
- benchmark-checklistSkill
- blast-radiusSkill
- codebase-designSkill
- correctSkill
- create-verification-skillSkill
- design-practiceSkill
- deslopSkill
- external-agent-handoffSkill
- howSkill
- nuclear-reviewSkill
- reflectSkill
- sharpenSkill
- tddSkill
- tech-diagramsSkill
- technical-writingSkill
- unslopSkill
- web-craftSkill
- whySkill
- writing-for-agentsSkill
The file
We split this AGENTS.md into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 33
Run scripts/sync-to-amp-skills.sh --dry-run to see what would change without committing.
≈ “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
Working in this repository
This repository is the public home of the skills (skills/<name>/). It is content only: no package
manager, no build step, no test runner. Verify a skill with python3 scripts/check-skills.py, which
fails on missing frontmatter, interface metadata, README rows, broken relative links, a
CREDITS.md without its LICENSE, a description that claims a sibling's trigger, or a
<!-- check:... --> tag that names no code its checker emits. Verify a figure with
python3 skills/tech-diagrams/scripts/self_check.py <source> --png <render>. When you write or edit
a SKILL.md here, follow skills/writing-for-agents/SKILL.md.
Three README landmarks for a new skill: the row at the end of the ## Skills table (README.md:51,
README.zh-CN.md:51), the tree entry at the end of ## Repository layout (README.md:268,
README.zh-CN.md:217), and the procedure in ## Adding a skill (README.md:326,
README.zh-CN.md:272). Re-run grep -n '^## ' README.md README.zh-CN.md if the line numbers
have moved.
Publishing a skill change
Amp serves personal skills from its own hosted Git repository, so a change here is not live in Amp
until it is copied there. After pushing a change under skills/ to main, run:
scripts/sync-to-amp-skills.shIt copies every skill (text files only - Amp serves no binaries), commits, and pushes to the
maintainer's personal skills repository. It reads origin/main, not the checkout it runs from, so
it is safe to run from an orb that is behind. It adds and updates; it never deletes, so a skill
created in Amp by hand survives a sync - and a skill deleted here has to be deleted there by hand.
The same script is what the daily schedule on the maintainer's thread runs, so this step is a
convenience for the common case, not the only path.
Run scripts/sync-to-amp-skills.sh --dry-run to see what would change without committing.
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.