jskills
AGENTS.md0 starsChanged 56 days agoMITView on GitHub
Keep or create a skill only when it captures knowledge or behavior that an agent cannot reliably infer on its own, such as: Do not create or retain a skill merely to restate general software-engineering knowledge. In particular, avoid: Prefer small, independent skills with explicit triggers and narrow responsibilities. Let the agent choose ordinary implementation steps. Constrain important outcomes with executable checks, keep durable project conventions in AGENTS.md, and use human or independent-agent review only where judgment is genuinely needed.
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 repository7 other files it gives its agents.
- ankifySkill
- anki-refineSkill
- code-kataSkill
- grill-meSkill
- md-code-tourSkill
- newsletterSkill
- task-assessorSkill
The file
We split this AGENTS.md into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 3
Keep or create a skill only when it captures knowledge or behavior that an agent cannot reliably infer on its own, such as:
- Distinctive: line 10
Do not create or retain a skill merely to restate general software-engineering knowledge.
- Distinctive: line 18
Prefer small, independent skills with explicit triggers and narrow responsibilities.
≈ “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 Selection
Keep or create a skill only when it captures knowledge or behavior that an agent cannot reliably infer on its own, such as:
- Exact personal conventions or output formats that must remain consistent.
- Specialized domain knowledge and validation procedures.
- Executable scripts, assets, or external protocols required to complete the task.
- A focused, repeatedly used workflow where omitted steps have caused concrete failures.
Do not create or retain a skill merely to restate general software-engineering knowledge. In particular, avoid:
- Long mandatory pipelines of planning, design, implementation, and status documents.
- Process artifacts that exist mainly to prove that a step was followed.
- Instructions already enforced more reliably by repository tests, linters, types, CI, or a short repository-level rule.
- Broad workflows that increase the user's coordination burden without producing measurably better results.
- Hidden dependencies on other skills or file conventions that are not intrinsic to the skill's purpose.
Prefer small, independent skills with explicit triggers and narrow responsibilities. Let the agent choose ordinary implementation steps. Constrain important outcomes with executable checks, keep durable project conventions in AGENTS.md, and use human or independent-agent review only where judgment is genuinely needed.
Evaluate a skill by whether it reduces repeated user effort and prevents observed failures, not by how comprehensive its procedure appears. If a skill no longer earns its maintenance and context cost, simplify or remove it.
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.