th3-skills
CLAUDE.md2 starsChanged 2 days agoMITView on GitHub
A collection of Claude Code skills in skills/, shipped as one plugin (.claude-plugin/) and installable per skill via npx skills. devops, qa, containers, agent-kit and documentation are umbrella skills with sub-skills in subdirs. Python runs through uv everywhere: locally, in CI, and in the repo's own scripts. No root test config. From inside the skill dir, set up once with the pins its workflow installs, then run pytest and the self-checks CI runs: Each skill has its own path-filtered workflow…
- Installs packages
In this repository23 other files it gives its agents.
- new-skillSkill
- agent-kit-costSkill
- agent-kit-prompt-cacheSkill
- agent-kitSkill
- containers-podmanSkill
- containersSkill
- dependency-vuln-explainerSkill
- design-coreSkill
- design-flowSkill
- devops-changeSkill
- devops-network-diagramSkill
- devops-release-scopeSkill
- devopsSkill
- documentation-doc-styleSkill
- documentation-pdf-operationsSkill
- documentationSkill
- poc-spec-driven-builderSkill
- qa-answer-reviewSkill
- qa-excel-importSkill
- qa-export-evidenceSkill
- qaSkill
- qa-ticket-reportSkill
- thai-mock-namingSkill
The file
We split this CLAUDE.md into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
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 10
Commit subjects are plain imperative, optionally scoped (qa: ..., fix(thai-mock-naming): ...).
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 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.
A collection of Claude Code skills in skills/, shipped as one plugin (.claude-plugin/) and installable per skill via npx skills. devops, qa, containers, agent-kit and documentation are umbrella skills with sub-skills in subdirs.
Git
- Branch
feat/<skill-slug>orfix/<slug>offdev; PR intodev;dev→mainis the release. - Commit subjects are plain imperative, optionally scoped (
qa: ...,fix(thai-mock-naming): ...).
Tests
Python runs through uv everywhere: locally, in CI, and in the repo's own scripts. No root test config. From inside the skill dir, set up once with the pins its workflow installs, then run pytest and the self-checks CI runs:
uv venv && uv pip install --no-build <pins from .github/workflows/<skill>.yml> # add --python 3.8 to test the oldest
uv run --no-project python -m pytest -q- qa: then
uv run --no-project python answer-review/scripts/qa.py selfcheck,… excel-import/scripts/run_web.py selftest,… ticket-report/scripts/ticket_report.py --selftest - devops: then
uv run --no-project python release-scope/scripts/release_scope.py --selftest - thai-mock-naming: pytest only (
pytest==8.3.5, the last pin that supports 3.8–3.13) - containers: then
uv run --no-project python podman/scripts/reclaim.py --help - agent-kit: no pytest;
uv run --script cost/scripts/agent_guard.py --selftestanduv run --script prompt-cache/scripts/pcache.py --self-test
Each skill has its own path-filtered workflow in .github/workflows/<skill>.yml, set up with astral-sh/setup-uv (uv supplies every matrix Python itself, so the runner's OS doesn't limit the versions). devops/qa/containers run Python 3.10–3.13; thai-mock-naming runs 3.8–3.13, so keep its code 3.8-compatible. CI installs pinned versions with uv pip install --no-build (wheels only, SonarCloud rules) — pin only versions that have wheels for every matrix Python, and quote the install line in YAML.
Agents
Spawning a subagent, fork or Workflow here follows skills/agent-kit/cost/SKILL.md: inline first, set model to the cheapest tier that can finish, every brief ends with how done is checked. A PreToolUse hook (agent_guard.py) adds the tier table to any Agent call that sets no model.
Skill conventions
- SKILL.md frontmatter is
nameanddescriptiononly. Descriptions are trigger-heavy ("Use WHENEVER ..."). Sub-skills are prefixed (qa-excel-import). - Scripts are stdlib-only Python unless the skill's CI already pins a dependency. Give them
--dry-run/--selftest/--helpwhere they change state or are testable. - devops and qa were generalised from a client project: no real hosts, IPs, company names or client data in any skill.
- Mock/sample data uses the thai-mock-naming markers (ทดสอบ นามสมมุติ), never real-looking common Thai names. The linter is
skills/thai-mock-naming/scripts/check_mock_names.py; a PostToolUse hook runs it on edited files. Files that must contain the blocklist go in that skill's.mocknamesignore. - A new skill needs a README entry and, if it has tests, its own workflow.
/new-skill <name>scaffolds this. - Every skill has an icon and a hero: add its drawing to
assets/icons.py(it refuses to build while a skill lacks one) and its steps toassets/skill_heroes.py(plus the index inassets/hero.py), then run them withuv run assets/…(each script declares its dependencies in a PEP 723 header) and check withuv run assets/check_heroes.py.
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.

