[English](TIPS.md) · [Português](TIPS.pt-BR.md)

# Tactical patterns

Short commands to operate the kits without relying on implicit memory. Adjust
paths to the emitted directory and read the plan before using `--apply`. Tips marked
*new in 2.6.0* need version 2.6.0 or later.

### 1. Inventory before creating

```bash
rg --files | rg '(^|/)(name|termo)'
```

### 2. Measure the Git tree without confusing untracked files

```bash
git status --short --branch
```

### 3. Separate local change from the last commit

```bash
git diff --stat && git diff --cached --stat
```

### 4. Find the decision, not only the implementation

```bash
rg -n -i 'decisão|decision|rejeitad|supersed' docs . --glob '*.md'
```

### 5. Run the preflight before the done gate

```bash
python frameworks/operator-kit-*/scripts/preflight.py --project .
```

### 6. Make the criterion fail before the fix

```bash
python -m pytest caminho/do/teste.py::test_caso -q
```

### 7. Validate Python syntax in isolation

```bash
python -m py_compile caminho/do/script.py
```

### 8. Run the artifact's self-test

```bash
python caminho/do/script.py --self-test
```

### 9. Close multiple criteria with AND

```bash
python frameworks/operator-kit-*/scripts/done_gate.py "python -m pytest -q" "python -m py_compile app.py"
```

### 10. Declare a partial without painting it green

```bash
python frameworks/operator-kit-*/scripts/done_gate.py "python -m pytest -q" --declare-partial "falta validar o destino externo"
```

### 11. Install a kit in plan mode first

```bash
python installers/kit-forge-*/kit_doctor.py install --kit <kit> --target . --host claude-code
```

### 12. Apply the installation only after reading the plan

```bash
python installers/kit-forge-*/kit_doctor.py install --kit <kit> --target . --host claude-code --apply
```

### 13. Install namespaced skills on Codex

```bash
python installers/kit-forge-*/kit_doctor.py install --kit <kit> --target . --host codex --apply
```

### 14. Verify the integrity of an emitted kit

```bash
python installers/kit-forge-*/kit_doctor.py verify <kit>
```

### 15. List registered installations

```bash
python installers/kit-forge-*/kit_doctor.py registry
```

### 16. Choose a checker from another provider

```bash
python multi-session/lane-kit-*/scripts/checker_router.py --maker claude --require
```

### 17. See the lanes without editing the registry

```bash
python multi-session/lane-kit-*/scripts/lane_board.py --help
```

### 18. Count items with an explicit ruler

```bash
python frameworks/operator-kit-*/scripts/live_count.py --help
```

### 19. Look for common silent errors

```bash
rg -n 'except\s+Exception|catch\s*\(|\.catch\(' --glob '*.py' --glob '*.js' --glob '*.ts'
```

### 20. Confirm the active GitHub account

```bash
gh api user --jq .login
```

### 21. Compare the remote before pushing

```bash
git ls-remote origin refs/heads/main
```

### 22. List MCPs per host

```bash
claude mcp list && codex mcp list
```

### 23. Generate the catalogue from the tree

```bash
python installers/kit-forge-*/tools/catalog_md.py . --write
```

### 24. Validate every published skill

```bash
find . -type d -name skills -exec python installers/kit-forge-*/tools/skill_lint.py --all {} \;
```

### 25. Record an end-to-end run as evidence (new in 2.6.0)

```bash
python -m hpp evidence run --id smoke-page --artifact out/report.html --artifact out/smoke.log -- python examples/evidence/smoke_page.py
```

### 26. Re-derive a bundle from the files on disk (new in 2.6.0)

```bash
python -m hpp evidence verify "$(ls -t .hpp/evidence/smoke-page-*.json | head -1)"
```

### 27. Measure a retriever before trusting it (new in 2.6.0)

```bash
python -m hpp retrieval eval examples/retrieval/suite.json --retriever-command '["python", "examples/retrieval/keyword_retriever.py"]'
```

### 28. Check that every citation resolves (new in 2.6.0)

```bash
python -m hpp cite check --text examples/citations/answer.md --context examples/citations/context.json
```

### 29. Measure a decider before trusting it (new in 2.6.0)

```bash
python -m hpp decide eval examples/typed-decisions/gotcha-family-suite.json --decider-command '["python", "examples/typed-decisions/baseline_decider.py"]'
```

### 30. Best-of-N is pass@N: select one, then still verify it (new in 2.6.0)

```bash
python multi-session/lane-kit-*/scripts/lane_board.py compete --task T-1 --items T-1a,T-1b --lane lead --model claude-opus
python multi-session/lane-kit-*/scripts/lane_board.py select --task T-1 --winner T-1a --lane review --model gpt-5 --reason "smaller diff, same tests"
```