Adonis0123/adonis-skills / skills/open-code-review-loop
open-code-review-loop
UnverifiedRun a bounded OCR-delegation review-fix-re-review loop on a Git workspace until current evidence validates NO_FINDINGS. Use only when the user wants verified findings fixed and independently re-reviewed until clean with codex, claude-code, grok-build, or cursor-cli. Do not use for one-shot open-code-review-delegate or read-only review. A full loop request asks at most once for an unnamed Reviewer (recommending a product other than the host), then runs host mutation checks without further confirmation. Fail closed on missing ocr, reviewer mutation, skipped files, malformed output, stale evidence, or exhausted rounds.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 19.3 KB |
| evals/evals.json | config | 20.8 KB |
| evals/trigger-eval.json | config | 2.5 KB |
| references/adapter-contracts.md | markdown | 13.5 KB |
| references/fix-contract.md | markdown | 2.7 KB |
| references/fix-schema.json | config | 857 B |
| references/review-contract.md | markdown | 3.1 KB |
| references/review-schema.json | config | 2.8 KB |
| scripts/build_review_bundle.pyexecutable | script | 27.0 KB |
| scripts/extract_product_output.py | script | 3.6 KB |
| scripts/test_extract_product_output.py | script | 6.2 KB |
| scripts/test_validate_fix_result.py | script | 5.1 KB |
| scripts/test_validate_round.py | script | 103.0 KB |
| scripts/validate_fix_result.py | script | 5.4 KB |
| scripts/validate_round.pyexecutable | script | 41.4 KB |
references/fix-contract.md
View on GitHub# Fixer result contract
Load this reference before invoking an external Fixer. Its user-authorized
isolated checkout must already be the authoritative `$REPO` used to build the
current evidence. Give the Fixer only the verified findings, frozen write
boundary, current evidence ID, and acceptance checks. The Fixer must return one
JSON object and no surrounding Markdown.
Use [fix-schema.json](fix-schema.json) when the selected product supports an
output schema.
```json
{
"schema_version": "1",
"evidence_id": "sha256:<current evidence>",
"finding_ids": ["OCR-001"],
"status": "FIXED",
"changed_paths": ["relative/path.ts"],
"summary": "Smallest coherent repair that was attempted",
"block_reason": null
}
```
For `BLOCKED`, provide a non-empty `block_reason`. For `FIXED`, use `null`.
After product-envelope extraction, validate the report before inspecting its
claims:
```bash
python3 <skill-dir>/scripts/validate_fix_result.py \
--fix "$ROUND_DIR/fix.json" \
--evidence-id "$CURRENT_EVIDENCE_ID" \
--finding-id "OCR-001" \
--output "$ROUND_DIR/fix-validation.json"
```
Malformed output gets at most one correction in the same known-delivery Fixer
session. A valid report still does not prove a mutation.
## Host-owned verification
The Fixer's `status`, `changed_paths`, summary, and claimed checks are
untrusted reports. They never prove that a write happened or that a finding is
resolved. After every Fixer return, the visible host must:
1. Recompute the real Git status and diff without resetting existing changes.
2. Reject every path outside the frozen write boundary as `HUMAN_GATE`.
3. Derive actual changed paths from Git rather than the Fixer response.
4. Re-open the cited code and run the host-owned acceptance checks.
5. Rebuild the OCR bundle before another Reviewer verdict can be accepted.
An external Fixer must operate in the user-authorized isolated writable `$REPO`
that the Reviewer and evidence builder have used since round 1. The checks
below are evidence gates after delivery; they are not a substitute for write
isolation and cannot safely undo an out-of-scope mutation in a dirty worktree.
This version has no patch-transfer mode.
If `status` is `FIXED` but Git/evidence did not change and the verified finding
still reproduces, classify the attempt as `FIXER_NO_MUTATION`. Delivery is
known in this case, so one correction in the same Fixer session is allowed.
If that correction also makes no valid mutation, return `UNVERIFIED`; do not
loop indefinitely.
A timeout, connection loss, non-zero exit, or truncated response after writes
may have occurred is different. Recompute Git state and return
`UNVERIFIED: DELIVERY_UNKNOWN_WITH_MUTATION`; never retry, reset, or assume the
reported `changed_paths` are complete.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |