collectiveai-team/botica / skills/review/review-refute
review-refute
VerifiedReview role that challenges findings against the diff, reachable entry points, reproductions and existing checks. Records contrary evidence and validates deterministic duplicates. Use after the other review-<role> skills ran, or when review-panel reaches its last step.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 10.7 KB |
| agents/openai.yaml | config | 103 B |
| references/REVIEW.md | markdown | 1.2 KB |
| references/output.md | markdown | 2.9 KB |
references/output.md
View on GitHub# Output of a review role
Every `review-<role>` skill writes the same JSON, so that `review-refute`, `review-panel` and later
`botica lint --reviewed` read one format. This file is copied into each role skill, and
`tests/test_skill_catalog.py` checks that the copies match.
## Posture
Review to find what is wrong, not to approve. Report only what the diff introduced or made
reachable. Report nothing that a deterministic check of the repository already reports: `botica
lint`, a hook of prek, a job of CI. Read the `## Do not report` section of the role file before
writing a finding.
## File
Each result also includes `inputs`, copied without modification from the fresh
`botica review plan` output. Use its full `inputs.base` and `inputs.head` as `base`
and `head`. `botica lint --reviewed <dir> --diff <target>` recomputes the plan and
rejects missing results or mismatched inputs. Short SHAs and branch names in the
example below are explanatory only and cannot pass evidence ingestion.
Review every branch and integration comparison of the plan, not only one merge-base diff.
Refute preserves each original finding. A dismissal adds `refutation_evidence` and
`refutation_references` (path, full revision, line, exact quote); `duplicate` also names
`duplicate_of`. Source quotes are checked against pinned Git objects. A `not-reproduced`
claim alone never dismisses a finding. `unreachable` and `pre-existing` also remain in the gate.
Only a duplicate with a retained target of no weaker blocking effect is omitted.
`.tmp/review/<short-head-sha>/<role>.json`:
```json
{
"role": "security",
"head": "2e12962",
"base": "main",
"verdict": "FAIL",
"findings": [
{
"id": "security-1",
"severity": "P1",
"path": "crates/botica-cli/src/server.rs",
"line": 42,
"title": "The token is compared with ==",
"evidence": "What in the code shows it, and the smallest case that breaks it.",
"reproduced": "yes | no | not-tried",
"how_to_verify": ["cargo test -p botica-cli server::token"],
"fix": "One sentence; no patch."
}
],
"checked": ["what this role reviewed"],
"not_checked": ["what it did not review, and why"],
"commands": [{"run": "cargo test -p botica-cli", "result": "pass"}]
}
```
## Verdict
- `PASS`: no finding at or above the blocking severity of the role file (P1 when it says none).
- `FAIL`: at least one finding at or above it.
- `INCONCLUSIVE`: the role could not do its review, for example no issue for `spec` or a test
suite that does not run. `not_checked` says why.
## Severity
- `P0`: breaks production, loses data or opens a security hole on the normal path.
- `P1`: wrong behaviour that a user or a caller reaches.
- `P2`: wrong on an edge case, or a gap that makes a later bug likely.
- `P3`: worth knowing; never blocks.
At most `max-findings` findings (10 when the role file says none), the most severe first.
`refute.json` holds every finding of the inputs and judges at most `max-findings` of them.
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: 472k | Updated | Claude CodeCodex | Unverified | 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 |