tkersey/fresh-eyes-blunder-pass

Fresh-Eyes Blunder Pass

Use this as a reusable second-pass falsifier. The pass objective is to reread the current artifact set as if another agent produced it, treating prior confidence as untrusted.

Modes

• implementation: new code, modified code, neighboring unchanged code, tests, imports, callsites, and invariants.
• closure: handoff packet, closure gates, verification evidence, negative-evidence gate, readiness claim, and reopen trigger.
• adjudication: act, address, validate-only, severity, direction, mutation approval, authority clearances, and handoff agenda rows.
• spec: authoritative brief, Evidence Brief, Gate Result, decision packet, proof bar, rollback, non-goals, requirement-to-test traceability, and execution handoff.
• plan: convergence claim, owners, dates, rollback, requirements, proof, scope, and execution order.
• skill-edit: trigger description, companion boundaries, validation requirements, stale paths, duplicate sections, false activation, and missed activation.
• negative-evidence: active exclusions, stale/reopened candidates, evidence anchors, applicability, overbroad bans, and reopening criteria.

Universal prompt

Carefully reread the artifact set with fresh eyes, looking for blunders, mistakes, errors, oversights, omissions, misconceptions, bugs, confusion, stale assumptions, scope creep, missing proof, and anything that would be embarrassing if it reached the user unchanged.

Required result

Return:

fresh_eyes_pass:
  mode: implementation | closure | adjudication | spec | plan | skill-edit | negative-evidence
  artifact_state_id: "..."
  checked_surfaces: []
  material_delta: yes | no
  findings: []
  fixes_or_required_updates: []
  verification_to_rerun: []
  fresh_eyes_delta: "none | summary"

If the pass changes the artifact, rerun the narrowest credible verification owned by the parent workflow. If it finds no material issue, report fresh_eyes_delta: none rather than inventing low-value findings.