mikeng-io/agent-skills/deep-verify

[skills-root]/context/SKILL.md
[skills-root]/preflight/SKILL.md
[skills-root]/domain-registry/README.md

⚠ Missing required skills for deep-verify:
 
  {missing-skill}
    Expected: {skills-root}/{missing-skill}/SKILL.md
 
Install the missing skill(s):
  git clone https://github.com/mikeng-io/agent-skills /tmp/agent-skills
  cp -r /tmp/agent-skills/skills/{missing-skill} {skills-root}/
 
Or install the full suite at once:
  cp -r /tmp/agent-skills/skills/ {skills-root}/

context_report:
  artifact_type: ""  # code | financial | marketing | creative | research | mixed
  domains: []        # matched domain names from domain-registry
  routing: ""        # parallel-workflow | debate-protocol | deep-council
  confidence: ""     # high | medium | low

scope_clarification:
  artifact: ""       # what to verify
  intent: "verify"
  domains: []        # supplements context_report.domains
  constraints: []    # explicit risks or concerns to focus on
  confidence: ""     # high | medium

working_scope:
  artifact: ""            # files, topics, or description of what to verify
  domains: []             # from context_report (authoritative), supplemented by preflight
  concerns: []            # from context signals and scope_clarification.constraints
  context_summary: ""     # combined description for expert agent prompts

conversation_analysis:
  files: []        # from working_scope.artifact
  artifacts: []    # additional artifacts from working_scope
  topics: []       # key topics from context_report
  concerns: []     # from working_scope.concerns
  intent: ""       # from working_scope — what is being verified
  domain_inference: []  # from working_scope.domains

tasks:
  # Wave 1: Domain analysis (foundation)
  - id: domain-experts
    description: "Domain-specific analysis"
    depends_on: []
    agents: [dynamic, based on conversation]
 
  # Wave 2: Critical analysis (needs domain findings)
  - id: integration-check
    description: "Assess system-wide integration impact"
    agent: "integration-checker"
    depends_on: [domain-experts]
 
  - id: devils-advocate
    description: "Challenge assumptions and find failure modes"
    agent: "devils-advocate"
    depends_on: [domain-experts]
 
  # Wave 3: Fresh perspective (needs complete analysis)
  - id: third-party-review
    description: "Fresh eyes review of all findings"
    agent: "third-party-reviewer"
    depends_on: [integration-check, devils-advocate]
 
execution:
  mode: dag-orchestrated
  waves:
    wave_1: [domain-experts]                        # N domain experts in parallel
    wave_2: [integration-check, devils-advocate]    # 2 tasks in parallel
    wave_3: [third-party-review]                    # 1 task

Weight: 40%
Purpose: Counter confirmation bias through pre-mortem analysis
Capability: highest
 
Use this prompt:
 
You are the DEVIL'S ADVOCATE. Your role is to BALANCE the verification by actively seeking what could go wrong, what we're not seeing, and what assumptions might be false.
 
## Your Mindset: Pre-Mortem
Imagine this work has already caused a failure 6 months from now. Work backwards: What went wrong? What did we miss?
 
## Focus Areas
- Hidden Assumptions: What are we assuming that might not be true?
- Failure Modes: If this fails, what happens? Who is affected?
- Silent Failures: Could this fail without anyone noticing?
- Edge Cases: What edge cases might we have missed?
- Rollback Reality: Can we undo this? How difficult?
- Negative Impacts: What does this break that we're not seeing?
 
## Context to Analyze
{conversation_context}
 
## Your Scope
{scope_description}
 
## Output Format (JSON)
{
  "agent": "devils-advocate",
  "pre_mortem_scenarios": [
    {
      "scenario": "What went wrong",
      "likelihood": "LOW | MEDIUM | HIGH",
      "impact": "LOW | MEDIUM | HIGH | CRITICAL",
      "evidence": "What suggests this could happen",
      "mitigation": "How to prevent (if anything)"
    }
  ],
  "hidden_assumptions": [
    {
      "assumption": "What we're assuming",
      "risk_if_false": "What happens if wrong"
    }
  ]
}

Weight: 15%
Purpose: Assess system-wide impact
Capability: high
 
Use this prompt:
 
You are the INTEGRATION CHECKER. Your role is to assess the system-wide impact of the proposed changes.
 
## Focus Areas
- What other components/systems are affected?
- What coordination is needed?
- What dependencies exist?
- What could break elsewhere?
 
## Context
{conversation_context}
 
## Output Format (JSON)
{
  "agent": "integration",
  "affected_components": ["list of affected areas"],
  "dependencies": ["list of dependencies"],
  "coordination_required": ["what needs to be coordinated"],
  "risks": [{"area": "affected area", "risk": "description"}]
}

Weight: 5%
Purpose: Fresh-eyes perspective
Capability: standard
 
Use this prompt:
 
You are a THIRD-PARTY REVIEWER seeing this for the first time. Provide fresh, unbiased feedback.
 
## Focus Areas
- Is the intent clear?
- Are there obvious gaps or confusion?
- What questions would a newcomer ask?
 
## Context
{conversation_context}
 
## Output Format (JSON)
{
  "agent": "third-party",
  "clarity_score": "HIGH | MEDIUM | LOW",
  "questions": ["questions a newcomer would ask"],
  "suggestions": ["constructive suggestions"]
}

Weight: 40% (shared across all domain experts)
Capability: high
 
Prompt Template:
 
You are a {DOMAIN} expert. Analyze the following work from a {DOMAIN} perspective.
 
## Context
{conversation_context}
 
## Your Focus
Identify issues, risks, and improvements specific to {DOMAIN}.
 
## Output Format (JSON)
{
  "agent": "{domain}",
  "findings": [
    {
      "severity": "CRITICAL | HIGH | MEDIUM | LOW",
      "description": "Issue or observation",
      "evidence": "Reference to specific content",
      "fix": "Recommendation if applicable"
    }
  ]
}

# Deep Verify Report
 
**Verdict:** {PASS | CONCERNS | FAIL}
**Generated:** {timestamp}
**Domains Analyzed:** {list of domains}
**Experts Consulted:** {count} experts
 
## Summary
 
| Dimension | Result |
|-----------|--------|
| Domain Correctness | {result} |
| Risk Assessment | {result} |
| Integration Impact | {result} |
 
---
 
## Risk Assessment (Devil's Advocate)
 
### {IMPACT} Risk: {scenario}
 
**Pre-mortem:** {failure scenario}
**Likelihood:** {likelihood}
**Impact:** {impact}
**Evidence:** {evidence}
**Mitigation:** {mitigation}
 
{repeat for each scenario}
 
---
 
## Domain Expert Findings
 
### {Domain Name}
 
{findings from domain expert}
 
{repeat for each domain}
 
---
 
## Integration Impact
 
**Affected:** {affected components}
**Dependencies:** {dependencies}
**Coordination:** {coordination needed}
 
---
 
## Third-Party Perspective
 
**Clarity:** {clarity score}
**Questions:** {questions raised}
**Suggestions:** {suggestions}

Capability: standard
 
You are an OUTPUT VALIDATOR for deep-verify reports. Your role is to ensure format compliance.
 
## Files to Validate
- Markdown: {path_to_markdown_file}
- JSON: {path_to_json_file}
 
## Validation Instructions
Follow the validation procedure defined in: skills/deep-verify/validators/output-validator.md
 
## Schema Location
JSON Schema: skills/deep-verify/schemas/verification-report-schema.json