tim-hub/powerball-harness / harness/skills/harness-review
harness-review
VerifiedMulti-angle code and plan review with security, scope, and UI profiles. Use when reviewing code, plans, PRs, or running pre-merge quality gates.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.2 KB |
| references/code-review.md | markdown | 2.1 KB |
| references/codex-closeout.md | markdown | 2.6 KB |
| references/codex-review.md | markdown | 1.2 KB |
| references/dual-review.md | markdown | 2.8 KB |
| references/governance.md | markdown | 2.2 KB |
| references/opencode-review.md | markdown | 2.3 KB |
| references/plan-review.md | markdown | 891 B |
| references/result-schema.md | markdown | 2.7 KB |
| references/scope-review.md | markdown | 344 B |
| references/security-profile.md | markdown | 11.8 KB |
| references/team-debate.md | markdown | 2.0 KB |
| references/ui-rubric.md | markdown | 7.3 KB |
| references/verdict-framework.md | markdown | 1.6 KB |
references/dual-review.md
View on GitHub# Dual Review (--dual)
Run Claude Reviewer and Codex Reviewer in parallel to improve review quality through different model perspectives.
## Prerequisites
- Codex CLI must be installed (verify with `"${CLAUDE_SKILL_DIR}/../../scripts/codex-companion.sh" setup --json`)
- If Codex is unavailable, falls back to Claude-only review
## Execution Flow
1. Check Codex availability
```bash
CODEX_AVAILABLE="$(bash "${CLAUDE_SKILL_DIR}/../../scripts/codex-companion.sh" setup --json 2>/dev/null | jq -r '.ready // false')"
```
2. Launch Claude Reviewer via the Task tool (standard review flow)
3. If Codex is available, launch `"${CLAUDE_SKILL_DIR}/../../scripts/codex-companion.sh" review` in parallel
```bash
# If BASE_REF is provided, specify --base. Use --json for structured output
bash "${CLAUDE_SKILL_DIR}/../../scripts/codex-companion.sh" review --base "${BASE_REF:-HEAD~1}" --json
```
4. Wait for both results
5. Verdict merge rules (evaluated in this order):
- Both APPROVE -> `APPROVE`
- Either is REQUEST_CHANGES -> `REQUEST_CHANGES` (adopt the stricter verdict)
- `critical_issues` are combined from both lists (no deduplication)
- `major_issues` are combined from both lists (no deduplication)
- `recommendations` are deduplicated and combined
## Output Format
Adds a `dual_review` field to the standard `review-result.v1` schema:
```json
{
"schema_version": "review-result.v1",
"verdict": "APPROVE | REQUEST_CHANGES",
"dual_review": {
"claude_verdict": "APPROVE | REQUEST_CHANGES",
"codex_verdict": "APPROVE | REQUEST_CHANGES | unavailable | timeout",
"merged_verdict": "APPROVE | REQUEST_CHANGES",
"divergence_notes": "Reason when verdicts diverge. E.g.: Claude detected a major issue in Performance, Codex found no issues"
},
"critical_issues": [],
"major_issues": [],
"observations": [],
"recommendations": []
}
```
### Special Values for `codex_verdict`
| Value | Meaning |
|-------|---------|
| `"unavailable"` | Codex CLI is not installed or unavailable |
| `"timeout"` | Codex review timed out (no response within 120 seconds) |
## Fallback
- **Codex unavailable**: Execute Claude-only and record `codex_verdict: "unavailable"`
- **Codex timeout**: Adopt Claude's verdict as-is and record `codex_verdict: "timeout"`
- **Codex review output invalid**: Treat as parse failure and record `codex_verdict: "unavailable"`
In all fallback cases, the Claude-only review result becomes the final verdict.
## Divergence Notes Format
When verdicts match (`claude_verdict == codex_verdict`), set `divergence_notes` to an empty string.
When verdicts diverge, record in the following format:
```
Claude: REQUEST_CHANGES (Security - SQL injection risk)
Codex: APPROVE (assessed the same location as no issue)
Adopted: REQUEST_CHANGES (prioritize the stricter verdict)
```
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |