mikeng-io/agent-skills/agent-council

{
  "participant": "{role-name}",
  "mode": "brainstorm",
  "proposals": [
    {
      "id": "P001",
      "title": "Short proposal title",
      "summary": "What this design does",
      "rationale": "Why it could work",
      "tradeoffs": [],
      "failure_modes": [],
      "open_questions": [],
      "recommended_when": "When to choose this proposal"
    }
  ],
  "assumptions": [],
  "questions": []
}

working_scope:
  artifact: "" # from context + preflight
  intent: "" # from context or preflight
  domains: [] # from context_report.domains (authoritative)
  constraints: [] # from preflight (empty if skipped)
  context_summary: "" # combined description for agent prompts
  intensity: "standard" # from routing signals

council_context:
  review_scope: "" # from working_scope.artifact
  context_summary: "" # from working_scope.context_summary
  artifact_type: "" # from context_report.artifact_type
  domains: [] # from working_scope.domains
  intensity: "standard" # from working_scope.intensity
  review_id: "" # Generate: agent-council-{YYYYMMDD-HHmmss}
  task_type: "review" # from working_scope.intent

Read: [skills-root]/domain-registry/domains/technical.md
Read: [skills-root]/domain-registry/domains/business.md
Read: [skills-root]/domain-registry/domains/creative.md

expert_count = len(domains) + 3  # domains + DA + IC + Synthesis Lead

{
  "expert_input": {
    "session_id": "{review_id}",
    "role": "{expert_role}",
    "scope": "{review_scope}",
    "task_type": "{task_type}",
    "context_summary": "{context_summary}",
    "domains": {domains},
    "intensity": "{intensity}",
    "focus_areas": "{role-specific focus from domain-registry}"
  }
}

You are a {domain} expert reviewing: {review_scope}
 
## Context
{context_summary}
 
## Your Domain Focus
{domain-specific focus areas from domain-registry}
 
## Your Task
Analyze the scope from your domain perspective. Return findings as JSON:
{
  "findings": [
    {
      "id": "{domain}-001",
      "severity": "CRITICAL | HIGH | MEDIUM | LOW | INFO",
      "title": "...",
      "description": "...",
      "evidence": "...",
      "remediation": "...",
      "domain": "{domain}",
      "confidence": "high | medium | low"
    }
  ],
  "domain_coverage": "{domain}",
  "assumptions_made": ["..."]
}
 
Be thorough. Challenge surface-level explanations. Look for root causes.

You are the DEVIL'S ADVOCATE for this review session.
 
## Context
{context_summary}
Scope: {review_scope}
Intensity: {intensity}
 
## Your Obligations (read debate-protocol/experts/devils-advocate.md for full protocol)
 
1. **MUST challenge every CRITICAL and HIGH finding** from primary reviewers
2. **SHOULD challenge MEDIUM findings** when you detect a pattern
3. **Cross-domain synthesis**: look for findings whose combination implies a worse issue
4. **Pre-mortem focus**: for each component, ask "what would cause this to fail in production?"
 
## Challenge Quality Standard
 
A valid challenge must either:
- (a) Identify a missing assumption that, if corrected, reduces severity
- (b) Propose an alternative explanation at least as plausible as the stated cause
- (c) Surface a scenario where the finding does not apply
 
Invalid challenges: "I don't think this is serious" (without mechanism).
 
## Output Format
 
{
  "challenges": [
    {
      "target_finding_id": "...",
      "challenge": "...",
      "proposed_severity": "CRITICAL | HIGH | MEDIUM | LOW | WITHDRAWN",
      "rationale": "...",
      "status": "pending"
    }
  ],
  "new_findings": [...],  // Failure modes you discovered through challenge
  "cross_domain_implications": [...]
}

You are the INTEGRATION CHECKER for this review session.
 
## Context
{context_summary}
Scope: {review_scope}
 
## Focus Areas (read debate-protocol/experts/integration-checker.md for full protocol)
 
1. **Interface mismatches**: where does component A assume something about B that isn't guaranteed?
2. **Undocumented contracts**: implicit dependencies that work by accident
3. **Error propagation gaps**: errors one component produces but callers don't handle
4. **Timing and ordering dependencies**: race conditions, initialization ordering
5. **Cross-cutting assumptions**: things that must be true globally but are only enforced locally
 
## Your Task
 
For each finding from primary reviewers: does it have cross-component implications beyond its stated scope?
If yes, surface those as integration findings.
 
## Output Format
 
{
  "integration_findings": [
    {
      "id": "INT-001",
      "severity": "...",
      "title": "...",
      "description": "...",
      "components_involved": ["..."],
      "implicit_contract": "...",
      "remediation": "...",
      "domain": "integration"
    }
  ],
  "cross_component_gaps": [...]
}

Task: primary-reviewer-{domain1}
Task: primary-reviewer-{domain2}
...
Task: devils-advocate
Task: integration-checker

You are a DEBATE COORDINATOR for expert council synthesis.
 
## Context
{context_summary}
Scope: {review_scope}
 
## Expert Outputs (from {N} experts)
{all expert outputs JSON}
 
## Your Task
 
1. **Collect Devil's Advocate challenges** against primary reviewer findings
2. **For each challenge**, determine outcome:
   - CONFIRMED: Finding holds after challenge
   - DOWNGRADED: Severity reduced (state new severity + rationale)
   - DISPUTED: Challenge has merit but finding not withdrawn
   - WITHDRAWN: Challenge reveals finding was invalid
 
3. **Collect Integration Checker findings** and merge with relevant primary findings
 
4. **Surface cross-expert synthesis**: findings that multiple experts independently surfaced
 
## Output Format
 
{
  "debate_round": 1,
  "confirmed_findings": [...],
  "downgraded_findings": [...],
  "disputed_findings": [...],
  "withdrawn_findings": [...],
  "integration_findings": [...],
  "multi_expert_confirmed": [...],  // Findings 2+ experts agreed on
  "challenge_notes": [...]
}

final_findings:
  multi_expert_confirmed: # 2+ experts independently surfaced
  single_expert: # One expert surfaced, survived challenge
  integration: # Cross-component implications
  disputed: # Challenge not resolved
  withdrawn: # Invalidated by challenge
 
verdict:
  FAIL:
    - Any CRITICAL confirmed by 2+ experts
    - 3+ HIGH confirmed findings
 
  CONCERNS:
    - 1-2 HIGH confirmed findings
    - Any disputed CRITICAL/HIGH
 
  PASS:
    - No CRITICAL/HIGH confirmed
    - Only MEDIUM/LOW/INFO