claudinho / rules
Cursor rule35 starsChanged 34 days agoMITView on GitHub
Scope a change and land it without a multi-round review loop
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .cursor/rules/bundle-bracket-pr.mdcCursor rule
- .cursor/rules/release-discipline.mdcCursor rule
- .cursor/rules/surface-parity.mdcCursor rule
- .cursor/rules/trust-boundary.mdcCursor rule
- claudinho-modeSkill
- verify-claudinhoSkill
- claudinho-modeSkill
- verify-claudinhoSkill
The file
Files
--- description: Scope a change and land it without a multi-round review loop globs: **/* alwaysApply: true --- # Change discipline The canonical text is **`AGENTS.md` → "Working agreement", "Validation scope", "Change discipline", and "Definition of Done"**. This file mirrors those rules for Cursor; keep it in sync. If the guides disagree, AGENTS.md wins. The short form: 1. **Complete authorized work through validation and self-review.** Resolve routine choices from the code; ask about missing information that changes scope or behavior. Plans, audits, and reviews stay read-only unless edits are requested. 2. **For user-facing features, write 3–5 acceptance criteria and explicit exclusions before implementation.** Put them in the PR description before the first commit; draft the description locally if there is no PR yet. 3. **Put a shared rule where every path reaches it, then delete the other copy** — not at the site the report happened to mention. 4. **Search for sibling paths**, including live/cache and statusline/hook readers. 5. **Make a new regression test fail before trusting it.** For call-site bugs, remove the call and confirm red; restore it afterward and preserve unrelated edits. 6. **Never assert wall-clock time.** Assert the observable consequence instead. 7. **Use AGENTS.md's validation scope and read the output.** Prose-only edits need `git diff --check`, link/contract checks, and the `node --input-type=module` block in [AGENTS.md's private-document boundary check](../../AGENTS.md#private-document-boundary-check). Code, dependency, and executable config changes need build → typecheck → test → lint plus relevant CI smoke/pack checks. `scripts/gate.sh` runs that list locally with one verdict per step (not `release:qa`, not CI's Node 22 floor, Windows and macOS jobs, which `scripts/push-and-watch.sh` reads on the pushed SHA); with `--commit <message-file>` it commits only when every step is ok, or, with `--allow-offline-audit`, when the audit alone was skipped offline (the commit line says so and the exit stays nonzero). The `verify-claudinho` skill and `node scripts/verify.mjs` are the offline inner loop; the control CLI replaces neither CI's jobs nor the live pass, `release:qa`. User-facing behavior and releases also need `release:qa` after build. Rerun checks for new changes, failures, or unresolved concerns; after a push, verify CI on that SHA. 8. **Start every task at the router skill**, `.claude/skills/claudinho-mode/SKILL.md` (mirrored at `.cursor/skills/claudinho-mode/SKILL.md`), which names the playbook for the task's kind, and find a failure class and the enforcer that refuses it in **`AGENTS.md` → "Rules and their enforcers"**.
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.

