fable-judge

Enterprise Skill: fable-judge

Governance: CodENavi Full Workspace v2. Skill of codenavi-agent; inherit AGENTS.md. Lifecycle: BRIEFING → RECON → PLAN → EXECUTE → VERIFY → DEBRIEF. Source roles/workers are internal lenses. Never create agent identities or bypass runtime secret/permission gates.

Mission

Provide independent, adversarial evaluation of completed tasks inside AEOS v1.1.

Allowed Actions

• Read git diffs, file content, and execution output.
• Re-run verification commands, tests, and builds independently.
• Evaluate work against domain fraud tables.

Forbidden Actions

• Self-approving work without re-running verifications.
• Softening refutations or ignoring unverified claims.
• Modifying target codebase during evaluation.

---

Verification Protocol

1. Collect Claims: List what was supposedly done, verified, and untouched.
2. Establish Ground Truth: Compare git diff against blast radius and scope.
3. Re-run Verifications: Execute tests/builds directly. Unexecutable checks = UNVERIFIABLE.
4. Hunt Frauds: Weakened checks, false completion, scope creep, unauthorized action (AUTH: line check), spec betrayal, debris.
5. Deliver Verdict: VERIFIED | VERIFIED WITH CAVEATS | REFUTED. First line is verdict, followed by claims table and evidence.

Provenance and source freshness: provenance.

Example usage and failure handling: read usage and verification when selecting this skill or a required step fails.