agentleFS
taecontroltaecontrol/skills/AGENTS.md

skills

AGENTS.md1 starsChanged 2 months agoMITView on GitHub

Agent skills, one folder each under skills/<bucket>/<name>/. README.md explains the buckets, the lifecycle, and versioning. Write every skill and this file with writing-for-agents, and finish with unslop. A skill changes because a real run went wrong. Adding a rule for each incident is what makes skills fail. Each new rule takes attention from the others, and the catalog grows until agents follow none of it closely. Change a skill this way instead: Each release bumps the version in package.json and…

The file

Files
AGENTS.md

taecontrol/skills

Agent skills, one folder each under skills/<bucket>/<name>/. README.md explains the buckets, the lifecycle, and versioning. Write every skill and this file with writing-for-agents, and finish with unslop.

Changing a skill

A skill changes because a real run went wrong. Adding a rule for each incident is what makes skills fail. Each new rule takes attention from the others, and the catalog grows until agents follow none of it closely. Change a skill this way instead:

  1. Start from the transcript of the run that went wrong, and find the point where the agent decided.
  2. Find why the text that should have caught it did not hold. The rule may be missing, but more often it exists and is too long, conflicts with another rule, sits in a skill the agent had not loaded at that point, or asks for something the model will not do from text alone. Fix that cause: delete, move, merge, or resolve the conflict. When a rule cannot hold as text, propose a mechanism to the owner, such as a hook or a line in their global instructions.
  3. One home per meaning. Each rule lives in the skill that owns that kind of work; other skills point to it. A change that copies a rule into a second skill moves it instead.
  4. Every added sentence fixes a cause. The goal is a skill that works, in as few words as it takes. Text that sits on top of a rule that failed, to cover the one case it missed, is a patch: change the rule that failed instead. A sentence earns its place when it fixes a cause from step 2 that no existing text covers. In the pull request, give the word counts before and after, and for each addition, the cause it fixes.
  5. Replay the real failure before merging. Give a fresh agent the changed text and the situation from the transcript just before the decision went wrong, without the outcome. The change holds when its decision differs from the one in the transcript. A scenario someone invented for the test does not count.

Each release bumps the version in package.json and package-lock.json, adds a short entry to CHANGELOG.md, and updates the bucket's README when a skill's one line changes.

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.