skills
AGENTS.md1 starsChanged 2 months agoMITView on GitHub
Use these conventions when adding or changing skills in this collection. Use templates/SKILL.md.template as a starting point, not a required outline. Improve instructions from demonstrated failures and remove rules that no longer help.
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 31,542 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 repository5 other files it gives its agents.
- frida-ios-hooksSkill
- frida-script-reviewSkill
- frida-tracing-discoverySkill
- frida-troubleshootingSkill
- ios-tweak-developmentSkill
The file
We split this AGENTS.md into the rules it gives an agent
10 rules. About 0 are common, 6 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 3
Use these conventions when adding or changing skills in this collection.
- Very specific: line 5
Put each skill in skills/<name>/SKILL.md.
- Distinctive: line 6
Keep the entry point short enough to read in one pass.
- Distinctive: line 7
Write for a capable agent.
- Distinctive: line 10
Keep instructions portable.
- Distinctive: line 12
Use primary sources for technical requirements.
- Distinctive: line 14
Do not turn a copied example's version numbers into a support promise.
- Distinctive: line 16
Test helpers locally. Distinguish syntax checks, mocked checks, model evaluations, and live-environment validation.
- Very specific: line 17
Keep the repository small: a short README, this file, the template, skills, and checks or sources that serve a concrete purpose.
- Very specific: line 19
Use templates/SKILL.md.template as a starting point, not a required outline.
≈ “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
Contributing skills
Use these conventions when adding or changing skills in this collection.
- Put each skill in
skills/<name>/SKILL.md. Use a lowercase, hyphenated name that matches the folder and a description explaining the task and when it applies. - Keep the entry point short enough to read in one pass. State the outcome, the decisions that matter, and how to verify the result.
- Write for a capable agent. Prefer specific failure modes and reasons over generic instructions to be careful or thorough.
- Move substantial conditional guidance into relative reference files. Say when to read each one. Avoid copying the same rules into several files.
- Include executable helpers only when they save repeated work or enforce a meaningful invariant. Give them explicit inputs, bounded effects, and runnable checks.
- Keep instructions portable. Use ordinary Markdown and common frontmatter; isolate host-specific features and provide a fallback when a named tool is unavailable.
- Preserve the user's task and existing authorization. A skill must not turn an explanation into an implementation or a local change into publication. Ask only for missing choices that materially affect the work.
- Use primary sources for technical requirements. Link the source near the claim; distinguish documented requirements, project conventions, and observations from one device or version.
- For imported skills, preserve the upstream license inside each independently installable folder and record its source commit and local changes in
THIRD_PARTY.md. Review upstream changes manually; keep optional companion routing usable when only one skill is installed. - Do not turn a copied example's version numbers into a support promise. Check current documentation and the actual project before prescribing toolchain or API details.
- Give new skills representative task prompts, a nearby task that should not trigger them, and observable pass criteria. Check decisions and artifacts, not exact phrasing.
- Test helpers locally. Distinguish syntax checks, mocked checks, model evaluations, and live-environment validation. Do not claim one proves the others.
- Keep the repository small: a short README, this file, the template, skills, and checks or sources that serve a concrete purpose. Add dependencies, CI, plugins, or publishing automation only when needed.
Use templates/SKILL.md.template as a starting point, not a required outline. Improve instructions from demonstrated failures and remove rules that no longer help.
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.

