BumpyClock/agent-templates / skills_archive/clawpatch
clawpatch
VerifiedClawpatch review/fix/revalidate loop. Verify findings, preserve IDs, triage with evidence.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.9 KB |
| references/cli.md | markdown | 1.8 KB |
| references/official-docs.md | markdown | 3.3 KB |
| scripts/revalidate_findings.py | script | 8.2 KB |
references/official-docs.md
View on GitHub# Clawpatch Docs Notes
Sources checked 2026-06-07: `https://clawpatch.ai/`, GitHub docs `index`, `quickstart`, `code-review`, `findings`, `patching`, `validation`, `configuration`, `safety`.
## Model
- Maps repo -> semantic features.
- Reviews bounded feature context -> findings.
- State in `.clawpatch/`: `config.json`, `project.json`, `features/`, `findings/`, `patches/`.
- `--json` = machine stdout; progress/warnings stderr.
## Common Flow
```bash
clawpatch init
clawpatch map
clawpatch review --limit 3 --jobs 3
clawpatch report
clawpatch fix --finding <id>
clawpatch revalidate --finding <id>
git diff
```
## Review
```bash
clawpatch review --limit 3
clawpatch review --limit 12 --jobs 4
clawpatch review --feature <id>
clawpatch review --since origin/main
clawpatch review --mode deslopify --limit 3
clawpatch review --provider codex --model <model>
```
- Consumes `clawpatch map` features.
- Claims feature locks; releases after review.
- `--since <ref>` limits to changed owned/context files vs `HEAD`.
- Rate limit: `--rate-limit-per-minute` or `CLAWPATCH_RPM`.
- Provider output must be strict JSON.
- Rejects unavailable evidence paths, stale line ranges, mismatched quotes.
## Deslopify
Simplification-only review. Report maintainability/performance findings rooted in accidental complexity.
Good targets: semantic duplication, pass-through wrappers, shadow modules, prod bloat, dead legacy paths, no-boundary defensive code, private-structure tests, type/build silencing, band-aid hacks.
Skip: file-size taste, normal framework boilerplate, broad architecture opinion.
## Findings
Finding fields: feature ID, title, category, severity, confidence, triage, evidence, reasoning, repro, recommendation, test gap, suggested regression test, minimum fix scope, status, linked patch attempts, history.
Statuses: `open`, `false-positive`, `fixed`, `wont-fix`, `uncertain`.
```bash
clawpatch status
clawpatch next
clawpatch show --finding <id>
clawpatch report --json
clawpatch triage --finding <id> --status false-positive --note "covered by contract test"
clawpatch revalidate --all --status open --limit 10
```
`triage` appends history; IDs stay stable.
## Fix/Validate
- `fix --finding <id>` only; explicit, scoped.
- Checks source worktree clean outside `.clawpatch/` when configured.
- Creates patch attempt, asks provider for plan/edit, runs configured `format`, `typecheck`, `lint`, `test`, records results, links attempt.
- Validation success -> finding `uncertain`; failure -> `open`.
- CLI does not mark `fixed` from patch pass alone. Revalidate after edits.
- Current limits: no `--skip-*` validation flags, no finding-targeted test generation, no parallel batch revalidation.
## Config/Safety
Config order: `--config`, `CLAWPATCH_CONFIG`, `$CLAWPATCH_STATE_DIR/config.json`, `clawpatch.config.json`, `.clawpatch/config.json`, defaults.
Env: `CLAWPATCH_STATE_DIR`, `CLAWPATCH_PROVIDER`, `CLAWPATCH_MODEL`, `CLAWPATCH_REASONING_EFFORT`.
Safe facts:
- `review`, `status`, `report`, `doctor`, `map --dry-run` do not edit source.
- `fix` needs `--finding <id>` and refuses dirty source by default.
- `.clawpatch/` state may change during runs.
- Codex provider review/revalidate uses read-only sandbox.
- No normal-flow auto commit/PR/landing/rollback/global process lock.
- After fix, caller owns `git diff` + tests before commit.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 318k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 318k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 318k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 318k | Updated | Claude CodeCodex | Verified | Install |