Kartik-Doye/Ravengard-AI-Recruiter / .agents/skills/referee
referee
VerifiedReviews a change against intent, design, correctness, complexity, security, tests, maintainability, and repository conventions. Use for code, diffs, pull requests, or implementation reviews that should produce only evidence-backed actionable findings.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.0 KB |
| references/finding-format.md | markdown | 4.5 KB |
| references/research.md | markdown | 805 B |
| references/review-protocol.md | markdown | 2.0 KB |
references/review-protocol.md
View on GitHub# Review protocol: scope a verdict to its evidence
Use this reference for a public contract, persistence, authorization,
concurrency, migration, deployment, or multi-package change. Ordinary bounded
diffs can use the entrypoint and finding format alone.
## Coverage record
Record acceptance criteria, diff range, files read, relevant callers or
consumers searched, tests and commands observed, repository instructions,
runtime surfaces considered, and exclusions. A verdict can be strong inside a
small scope; it must not imply coverage outside that scope.
## Impact map
For each changed boundary, identify only applicable items:
| Boundary | Check |
| --- | --- |
| Public API or event | clients, serialization, docs, versioning, compatibility |
| Data or migration | existing rows, backfill, rollback, ordering, deployment sequence |
| Authorization | identity source, object access, failure disclosure, audit trail |
| Async or retry path | idempotency, timeout, cancellation, duplicate effects, observability |
| Configuration | defaults, overrides, secrets handling, manifests, safe fallback |
| Dependency upgrade | lockfile, supported range, generated code, breaking behavior |
## Evidence policy
State fact when source, diff, test, or reproduction demonstrates it. State
inference when evidence supports but does not prove it. State unknown when the
needed caller, runtime, or configuration is unavailable. Severity follows
consequence and confidence together: a high-consequence possibility can merit
investigation, but should name confirmation needed.
## Review recovery
If criteria, complete diff, history, test output, or consumer code is missing,
list exact missing evidence and its effect on verdict. Do not fill the report
with generic advice. Resume only after evidence arrives or issue a bounded
review that labels the exclusion.
## Root-cause grouping
When locations fail for one reason, report root cause first, affected locations,
and one repair direction. Separate findings only when their fixes or acceptance
consequences differ.
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: 194k | Updated | Claude CodeCodex | Verified | Install |