cv-kit
AGENTS.md1 starsChanged 23 days agoMITView on GitHub
How to change cv-kit without breaking the plugin or its prompts. - skills/*/SKILL.md and agents/*.md are the prompts. Every word is read on each use, so a change that adds words must earn them. - hooks/concise.md and hooks/progress.md are injected into every session (and concise.md into every subagent). tests/test_concise_hook.py caps them at 2,200 and 1,200 characters. - rules/ is read by develop from the plugin. A plugin cannot ship rules directly, so scripts/install_rules.py copies them into a project's .claude/rules/ when…
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 32,671 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.
The file
We split this AGENTS.md into the rules it gives an agent
6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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 19
Test first. Change or add the test in tests/, watch it fail, then change the prompt or script.
- Very specific: line 20
Keep the README skill table, evals/README.md coverage and the files in sync.
- Very specific: line 36
Run scripts/run_evals.sh before a release.
- Very specific: line 43
Use one term for one thing, and must for a requirement.
- Distinctive: line 44
Keep exact identifiers, commands, paths, numbers and quotes.
- Distinctive: line 45
No em or en dashes in prose, no bold label on every list item, no closing line that repeats the paragraph.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 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
Guide for agents and contributors
How to change cv-kit without breaking the plugin or its prompts.
What is where
skills/*/SKILL.mdandagents/*.mdare the prompts. Every word is read on each use, so a change that adds words must earn them.hooks/concise.mdandhooks/progress.mdare injected into every session (andconcise.mdinto every subagent).tests/test_concise_hook.pycaps them at 2,200 and 1,200 characters.rules/is read bydevelopfrom the plugin. A plugin cannot ship rules directly, soscripts/install_rules.pycopies them into a project's.claude/rules/when the user asks.shared/holds the catalogues the prompts point at.shared/prose.mdis the one source for writing style; the regex part of it lives inPROSE_TELLSinscripts/lint_research.py. Change both together.evals/holds the trigger and behaviour cases;evals/README.mdexplains the format.
Rules for changes
- Test first. Change or add the test in
tests/, watch it fail, then change the prompt or script. - Keep the README skill table,
evals/README.mdcoverage and the files in sync;scripts/validate_plugin.pychecks the README table. - A new or removed skill needs its fire and no-fire eval cases added or removed;
tests/test_evals.pychecks it. - A new regex grader needs sample strings in
REGEX_SAMPLES. - Credit any project an idea came from in the README's Acknowledgements, with its licence. Write the rule in our own words; copy nothing from a non-commercial source.
- Commits follow
rules/commit-vcs.md.
Checks before publishing
.venv/bin/python scripts/validate_plugin.py
.venv/bin/python -m pytest tests -q
claude plugin validate . --strictRun scripts/run_evals.sh before a release; it calls the API and costs money.
Writing style
Prompts, docs, comments and messages follow shared/prose.md. In short:
- Lead with the main point. Use common words, active voice, short sentences.
- Use one term for one thing, and
mustfor a requirement. - Keep exact identifiers, commands, paths, numbers and quotes.
- No em or en dashes in prose, no bold label on every list item, no closing line that repeats the paragraph.
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.

