consensus-engine

name: consensus-engine description: Multi-validator agreement gate. N independent validators against same feature; synthesize confidence-scored verdict. triggers: - "consensus validation" - "validate with N reviewers" - "agreement gate" - "high-confidence validation" - "ship gate"

consensus-engine

Execution-time agreement gate. Spawns ≥2 (default 3) independent validators against the same feature; synthesizes confidence-scored verdict.

Behavior contract

1. Spawn N consensus-validator subagents in parallel via Task tool. Each owns an isolated evidence subdirectory:

   e2e-evidence/consensus/<run-id>/
     validator-1/
     validator-2/
     validator-3/
     report.md  (synthesizer-owned)
   

2. Each validator runs the full journey list independently. Blind to peer verdicts.
3. Synthesize via consensus-synthesis (folded into this skill):
   • Per journey: compute (pass_count, fail_count, total).
   • Map to synthesis state:

     State	Condition	Verdict	Confidence
     UNANIMOUS_PASS	All PASS	PASS	HIGH
     UNANIMOUS_FAIL	All FAIL	FAIL	HIGH
     MAJORITY_PASS	≥⅔ PASS	PASS (after disagreement protocol)	MEDIUM
     MAJORITY_FAIL	≥⅔ FAIL	FAIL (after disagreement protocol)	MEDIUM
     SPLIT	Neither side ⅔	DISAGREEMENT_UNRESOLVED	LOW

4. Disagreement protocol (only on non-unanimous):
   • Identify diverging criteria.
   • Invoke sequential-analysis on the divergent evidence.
   • Re-resolve via case (a) re-run missing validator, (b) re-run minority, (c) sharpen criterion (escalate to planner), (d) discard erroring validator.
   • If unresolvable → SPLIT with LOW confidence; escalate.

File ownership (LOAD-BEARING)

• Coordinator: writes NOTHING
• Validator-N: writes ONLY to validator-<N>/ subdirectory
• Synthesizer: writes ONLY report.md

A validator that writes outside its directory invalidates the run.

When to use

• High-stakes change (security, payments, production data)
• Pre-major-release validation
• After a near-miss bug to harden the gate
• User explicitly requests --mode consensus

When NOT to use

• Routine refactor with no behavior change
• Exploratory work
• Speed-critical iteration (consensus is 3× the wall time of single-validator)

Iron rules

• Confidence never upgrades — a late-arriving PASS verdict doesn't turn SPLIT into UNANIMOUS_PASS.
• Evidence quality cannot substitute for agreement.
• SPLIT is a real, reportable outcome. Do NOT silently downgrade to MAJORITY.