skills
CLAUDE.md1 starsChanged 3 days agoMITView on GitHub
AGENTS.md holds the marketplace layout, manifest rules and writing style. This file adds what it does not cover. archive/ holds retired plugins (archive/tutor/, replaced by study in kadeski/study). Nothing there ships or is listed in a marketplace, and scripts/check.sh skips it. Don't edit it; to bring a plugin back, move it to plugins/ and list it again. Claude-only, with no version on purpose (see AGENTS.md). Its tests in tests/sdlc/ are real headless model runs that cost money, so run them…
The file
We split this CLAUDE.md into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
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 32
Never read evals/<skill>/test/ or evals/results/*/test/ while editing a skill, and never put case content into a skill.
- Distinctive: line 33
After changing cases or graders, make a new baseline.
≈ “Common” is an estimate: at least 1 in 5 of the 28,029 CLAUDE.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
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
AGENTS.md holds the marketplace layout, manifest rules and writing style. This file adds what it does not cover.
Commands
scripts/check.sh- run before every commit. Needsjqand theclaudeCLI. It checks that each manifest is valid and named for its folder, that the marketplace lists every plugin, that eachSKILL.mdnamematches its folder, that no plugin file has a credential-like name (key, token, secret, password, auth) inside a${...}substitution, which the plugin directory holds underMCP_FORWARDS_CREDENTIAL_ENV, and runsclaude plugin validate --stricton every plugin and the repo. The one warning it lets through is a plugin with noversion. It also warns, without failing, when a plugin's tracked files differ from origin/main (as last fetched) but its version does not..github/workflows/check.ymlrunsscripts/check.shon every push and PR, so it holds in a fresh clone or a Projects thread where the pre-commit hook is off.scripts/hooks/pre-commitrunsscripts/check.shand blocks the commit if it fails. Turn it on once per clone withgit config core.hooksPath scripts/hooks(this covers every worktree). It checks the working tree, so unstaged changes count too.git commit --no-verifyskips it.- Evals (see
evals/README.md):evals/run.py dry <skill> -v- grade untouched fixtures, no agent or judge, free.evals/run.py run <skill> --split train --case <glob> --runs 1- one case.evals/run.py compare <old results> <new results>- only compare runs made with the same--modeland--judge-model.
/hillclimb <skill>is a project skill in.claude/skills/hillclimb/.
Archive
archive/ holds retired plugins (archive/tutor/, replaced by study in kadeski/study). Nothing there ships or is listed in a marketplace, and scripts/check.sh skips it. Don't edit it; to bring a plugin back, move it to plugins/ and list it again.
The sdlc plugin
Claude-only, with no version on purpose (see AGENTS.md). Its tests in tests/sdlc/ are real headless model runs that cost money, so run them by hand. Each file's header says what it needs. slice-pr.sh opens a PR in kadeski/sdlc-scratch.
.claude/skills/ holds symlinks to every skill in plugins/sdlc/skills/ and plugins/doc-lint/skills/, because Projects threads load a cloned repo's .claude/skills/ but no plugins. A new skill gets a symlink too (ln -s ../../plugins/<plugin>/skills/<name> .claude/skills/<name>); scripts/check.sh fails when one is missing.
Skills read each other as files (../verify/SKILL.md), so keep them as siblings in skills/.
Evals
- Never read
evals/<skill>/test/orevals/results/*/test/while editing a skill, and never put case content into a skill. - After changing cases or graders, make a new baseline; old scores no longer compare.
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.

