heinzel
CLAUDE.md2 starsChanged 14 days agoMITView on GitHub
Plan, delegate, verify, report outcomes. Ask the user only for product direction, anything irreversible or outward-facing (publishing, deleting, force-pushing, spending), credentials, and conflicts between their instructions. Never ask them to pick models, agents or tools. Route work to the cheapest tier that does it well, cheap to expensive: Keep each agent's frontmatter model and effort unless a skill says otherwise. Run independent pieces (no shared results or files) concurrently in background subagents; keep the piece that needs this conversation. Run…
In this repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .github/AGENTS.mdAGENTS.md
- tests/AGENTS.mdAGENTS.md
- .github/CLAUDE.mdCLAUDE.md
- starter/CLAUDE.mdCLAUDE.md
- template/CLAUDE.mdCLAUDE.md
- tests/CLAUDE.mdCLAUDE.md
- add-moduleSkill
- add-operationSkill
- codifySkill
- raise-the-barSkill
- releaseSkill
- reviewSkill
- scaffold-pluginSkill
- self-auditSkill
- ship-smallSkill
- write-adrSkill
- write-testsSkill
The file
We split this CLAUDE.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 9
Never ask them to pick models, agents or tools.
- Distinctive: line 25
Keep each agent's frontmatter model and effort unless a skill says otherwise.
- Distinctive: line 27
Run independent pieces (no shared results or files) concurrently in background subagents.
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 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
@AGENTS.md
Claude Code
Plan, delegate, verify, report outcomes. Ask the user only for product direction, anything irreversible or outward-facing (publishing, deleting, force-pushing, spending), credentials, and conflicts between their instructions. Never ask them to pick models, agents or tools.
Route work to the cheapest tier that does it well, cheap to expensive:
| Work | Who |
| Anything a make target or script does | the tool, not a model |
| Why CI or a check failed | ci-triage |
| Sweeping searches across many files | built-in Explore |
| Tests, CLI/DX, docs, CI/CD review | test-engineer, dx-reviewer, docs-reviewer, release-engineer |
| Correctness and security review | critic, security-reviewer (skill review picks the reviewers) |
| Design decisions, ADRs, structural review | architect; a whole area before a release: skill raise-the-bar |
| Whole-repo consistency audit, weekly or before a release | consistency-auditor; the loop: skill self-audit |
| What did the maintainer have to point out | correction-miner |
| Implementation | this session |
Keep each agent's frontmatter model and effort unless a skill says otherwise.
Run independent pieces (no shared results or files) concurrently in background subagents; keep the piece that needs this conversation. Run research, measurement and review alongside implementation. A subagent that edits the repository gets its own worktree unless it is the only writer. Check each result before building on it.
Hooks (.claude/settings.json) lint the Python you write, block --no-verify, --force, pull requests not
targeting main and checks piped into a push, and refuse to end a turn while make check-fast fails. Fix the cause;
never work around a hook.
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.