LaizyIO avatar

workflow-challenger

Critical review and gap analysis skill that can be invoked at any workflow stage. Use to challenge d

作者 LaizyIO|オープンソース

Workflow Challenger Skill

Purpose

Act as a critical reviewer and devil's advocate at any stage of the feature workflow. This skill performs deep analysis to:

  • Challenge assumptions and decisions
  • Identify gaps and missing specifications
  • Surface unanswered questions
  • Verify coherence between documents and codebase
  • Ensure nothing has been overlooked

Key principle: Better to surface issues early than discover them during implementation or testing.

When to Use This Skill

Use this skill:

  • After completing CDC.md to verify specification completeness
  • After research to challenge technical decisions
  • After planning to verify plan coherence with requirements
  • Before implementation to catch missing details
  • After test failures to understand root causes
  • Anytime something feels incomplete or uncertain
  • When user wants a second opinion on approach

IMPORTANT: User Interaction

Use the request_user_input tool when clarification is needed on identified gaps.

After analysis, use request_user_input to:

  • Confirm critical issues need resolution
  • Get user input on ambiguous findings
  • Prioritize which gaps to address first
request_user_input:
  questions:
    - question: "I found that the CDC doesn't address error handling. How should we proceed?"
      header: "Gap Found"
      options:
        - label: "Add to CDC now"
          description: "Update the specification before continuing"
        - label: "Note for later"
          description: "Document as known gap, address during implementation"
        - label: "Not needed"
          description: "This case won't occur in practice"
      multiSelect: false

This ensures:

  • User is aware of identified issues
  • Clear decision-making on how to proceed
  • Gaps are explicitly acknowledged or resolved

Analysis Sources

The challenger analyzes multiple sources to build comprehensive context:

1. Project Documentation

Look for documentation folders at project root:
- [DOC]-{ProjectName}/
- docs/
- documentation/
- .doc/
- wiki/

Read all relevant documentation to understand:

  • Project architecture
  • Business rules
  • Existing conventions
  • Historical decisions

2. Codebase Analysis

Analyze existing code to understand:
- Current implementation patterns
- Data models and schemas
- API contracts
- Service dependencies
- Configuration patterns

3. Workflow Deliverables

Review current workflow documents:
- CDC.md (specification)
- findings.md (research)
- Plan.md (implementation plan)
- test-plan.md (test strategy)

4. Obsidian/Knowledge Base

If Obsidian MCP is available:
- Search for related notes
- Find linked concepts
- Discover historical context

Challenge Workflow

Phase 1: Context Gathering

  1. Identify current stage - What deliverable is being challenged?
  2. Read the deliverable - Understand what has been documented
  3. Explore project docs - Find [DOC]-* folders and read relevant files
  4. Analyze codebase - Understand existing patterns and constraints
  5. Build mental model - Synthesize all information
Example context gathering:
1. Read CDC.md for "email notifications" feature
2. Found [DOC]-Backend/ with architecture.md
3. Found existing NotificationService in codebase
4. Found EmailTemplate component already exists
5. Mental model: Feature partially exists, CDC doesn't mention this

Phase 2: Gap Analysis

Systematically check for gaps in each category:

2.1 Missing Requirements

  • Are all user stories covered?
  • Are edge cases addressed?
  • Are error scenarios defined?
  • Are non-functional requirements specified?

2.2 Unclear Specifications

  • Are there ambiguous terms?
  • Are there assumptions not validated?
  • Are there "TBD" or placeholder sections?
  • Are acceptance criteria measurable?

2.3 Incoherence Detection

  • Does the document contradict itself?
  • Does it contradict project documentation?
  • Does it contradict existing codebase?
  • Does it contradict other workflow documents?

2.4 Missing Context

  • Is existing functionality acknowledged?
  • Are dependencies identified?
  • Are integration points clear?
  • Is the scope realistic?

See references/gap-analysis-guide.md for detailed checklist.

Phase 3: Challenge Formulation

For each identified issue, formulate a clear challenge:

Structure:

## Challenge: [Title]

**Type:** Gap / Ambiguity / Incoherence / Missing Context
**Severity:** Critical / Major / Minor
**Location:** [Document section or code location]

**Observation:**
[What was found or missing]

**Question:**
[Specific question to resolve the issue]

**Recommendation:**
[Suggested resolution or investigation]

Example:

## Challenge: Existing Email System Not Addressed

**Type:** Missing Context
**Severity:** Critical
**Location:** CDC.md Section 4 (Functional Requirements)

**Observation:**
The CDC specifies creating an EmailService, but the codebase already
has `src/services/NotificationService.ts` with email capabilities.
The CDC doesn't mention whether to extend, replace, or ignore it.

**Question:**
Should we extend the existing NotificationService or create a
separate EmailService? What happens to existing email functionality?

**Recommendation:**
Update CDC to explicitly address the existing NotificationService.
Consider extending it rather than creating parallel functionality.

Phase 4: Coherence Verification

Cross-check between documents and reality:

CDC vs Codebase:

  • Are referenced components real?
  • Do proposed integrations make sense?
  • Are technology choices consistent?

CDC vs Project Docs:

  • Does it follow documented architecture?
  • Does it respect documented constraints?
  • Is terminology consistent?

Findings vs CDC:

  • Does research address all CDC requirements?
  • Are POC results aligned with specifications?
  • Are chosen patterns appropriate?

Plan vs Findings:

  • Does plan implement research recommendations?
  • Are dependencies correctly sequenced?
  • Is scope consistent?

Phase 5: Report Generation

Compile findings into a structured report:

# Challenge Report: [Deliverable Name]

**Date:** [Date]
**Stage:** [Specification / Research / Planning / etc.]
**Challenger:** Codex

## Summary

[Brief overview of findings]

- **Critical Issues:** X
- **Major Issues:** X
- **Minor Issues:** X

## Critical Issues

[Must be resolved before proceeding]

## Major Issues

[Should be resolved, but not blocking]

## Minor Issues

[Nice to address, low impact]

## Coherence Check

| Source A | Source B | Status | Notes |
|----------|----------|--------|-------|
| CDC | Codebase | ⚠️ | [Issue] |
| CDC | Project Docs | ✅ | Aligned |

## Questions for User

1. [Question 1]
2. [Question 2]

## Recommendations

1. [Recommendation 1]
2. [Recommendation 2]

## Conclusion

[Overall assessment and suggested next steps]

Challenge by Stage

Challenging CDC.md (Specification)

Focus on:

  • Requirement completeness
  • Scope clarity
  • Acceptance criteria measurability
  • Alignment with project architecture
  • Existing functionality conflicts

Key questions:

  • "Is this already partially implemented?"
  • "Does this conflict with existing features?"
  • "Are all personas addressed?"
  • "What happens in failure scenarios?"

Challenging findings.md (Research)

Focus on:

  • Technical decision rationale
  • Alternative evaluation completeness
  • POC validity
  • Integration feasibility
  • Risk identification

Key questions:

  • "Why this approach over alternatives?"
  • "Was the POC representative?"
  • "Are performance implications understood?"
  • "What could go wrong?"

Challenging Plan.md (Implementation)

Focus on:

  • Step granularity appropriateness
  • Dependency correctness
  • Parallelization opportunities
  • Missing steps
  • Validation criteria clarity

Key questions:

  • "Is this step too vague to implement?"
  • "What if step X fails?"
  • "Can these steps run in parallel?"
  • "How do we know this step is complete?"

Challenging test-plan.md (Testing)

Focus on:

  • Coverage completeness
  • Test priority appropriateness
  • Edge case coverage
  • Performance test inclusion
  • Realistic test scenarios

Key questions:

  • "Are negative scenarios tested?"
  • "Is this test redundant with another?"
  • "What's the most likely failure mode?"
  • "Are integration points tested?"

Context Discovery Patterns

Finding Project Documentation

# Common documentation locations
Glob: [DOC]-*/**/*.md
Glob: docs/**/*.md
Glob: documentation/**/*.md
Glob: *.md (root level)
Glob: .github/**/*.md

Finding Related Code

# Based on feature name/domain
Grep: "feature-keyword"
Glob: **/*Service*.ts
Glob: **/*Controller*.cs
Glob: **/models/**/*

Using Obsidian MCP

If Obsidian available:
- Search notes for feature keywords
- Find linked architecture decisions
- Discover related past implementations

Challenge Intensity Levels

Quick Review (5-10 min)

  • Scan document structure
  • Check for obvious gaps
  • Verify key sections present
  • Spot-check coherence

Standard Review (15-30 min)

  • Full document read
  • Cross-reference with codebase
  • Check project documentation
  • Formulate key questions

Deep Review (30-60 min)

  • Comprehensive analysis
  • Full codebase exploration
  • All documentation reviewed
  • Detailed challenge report

Tips for Effective Challenging

  1. Be constructive - Goal is improvement, not criticism
  2. Be specific - Vague concerns aren't actionable
  3. Prioritize - Focus on critical issues first
  4. Suggest solutions - Don't just identify problems
  5. Consider context - Time/resource constraints matter
  6. Trust but verify - Check claims against reality
  7. Document findings - Create actionable report
  8. Ask, don't assume - Formulate questions when uncertain

Integration with Workflow

The challenger can be invoked:

  • Explicitly: User requests challenge review
  • At transitions: Between workflow phases
  • On demand: Anytime during development
Phase 0: Specification → [Challenge CDC] → Phase 1: Research
Phase 1: Research → [Challenge Findings] → Phase 2: Planning
Phase 2: Planning → [Challenge Plan] → Phase 3: Implementation
...

Bundled Resources

  • references/challenge-checklist.md - Comprehensive checklist by stage
  • references/gap-analysis-guide.md - Detailed gap analysis methodology

Compatibility Mapping (Claude -> Codex)

  • AskUserQuestion -> request_user_input
  • Task tool with subagent_type -> spawn_agent with agent_type
  • Glob/Grep/Read instructions -> use shell tooling (rg, Get-Content, Get-ChildItem)
  • If a Claude-specific primitive is unavailable, use the nearest Codex-native tool with equivalent intent

UX DesignOps Integration

When challenging UI work, verify:

  • CDC and FEAT docs state user goal, primary task, states, accessibility, and responsive constraints;
  • findings include UX/UI research when the feature is user-visible;
  • plans include visual verification and UX doc sync;
  • implementation preserves business logic and design system constraints;
  • tests include state matrix, desktop/mobile, keyboard/focus, and accessibility basics;
  • [DOC]-*/11-UX-DesignOps/, MOC-UX.md, and DESIGN.md are updated when required.

Flag any no-rendering UI validation as a blocker.