/exit-criteria

Exit Criteria - Complete Checklists

After Phase 1: Discovery (Primary Agent)

Before proceeding to Phase 2, verify:

• Explored codebase thoroughly

  • Checked relevant files with Read/Glob/Grep
  • Reviewed existing patterns and conventions
  • Examined similar implementations
  • Checked docs/, README.md, CLAUDE.md for context
  • Reviewed recent git commits for development focus

• Requirements explicitly clarified

  • Used AskUserQuestion to fill gaps codebase couldn't answer
  • Asked questions one at a time (not bundled)
  • Preferred multiple choice questions when possible
  • Clarified purpose, constraints, and success criteria

• Mental model built

  • Clear understanding of what's being built
  • Known constraints and non-functional requirements
  • Success criteria defined
  • Relevant existing patterns identified

• Ready for option analysis

  • Have enough context to propose viable approaches
  • Know which existing patterns to follow or adapt
  • Understand trade-offs that matter for this project

After Phase 2: Option Analysis

Before creating design document, verify:

• At least 2 options compared with trade-offs

  • OR clear "No Alternatives" rationale provided
  • Options grounded in codebase reality, not abstract possibilities
  • Trade-offs explained (complexity, maintainability, performance, etc.)
  • Referenced specific files/patterns from codebase

• User approval received

  • Presented options conversationally (not formal tables)
  • Led with recommended option and reasoning
  • Used AskUserQuestion to confirm chosen approach
  • User explicitly approved the direction

• Respected existing architecture

  • Proposals align with established patterns
  • Use existing libraries and frameworks when possible
  • Follow project's architectural style

• Ready to create design document

  • Have complete context from Phase 1
  • User has approved chosen approach
  • Know which files and patterns to reference

After Phase 3: Design Creation

Before proceeding to Phase 4 (Reflection), verify:

• Core sub-agents completed

  • Architecture Research sub-agent completed with recommendations
  • Best Practices Research sub-agent completed with BDD scenarios
  • Context & Requirements Synthesis sub-agent completed with requirements

• Additional sub-agents completed (if launched)

  • Each additional sub-agent completed successfully
  • All specialized research incorporated into design

• Results integrated successfully

  • Conflicts between sub-agent findings resolved
  • Consistent unified design document created
  • All findings incorporated into design

• Design document structure created

  • Folder pattern: docs/plans/YYYY-MM-DD-<topic>-design/
  • _index.md created with Context, Requirements, Rationale, Detailed Design
  • bdd-specs.md created with BDD scenarios
  • architecture.md created with architecture details
  • best-practices.md created with best practices and considerations

• Design documents include required sections

  • Context and discovery results
  • Finalized requirements and success criteria
  • Rationale for chosen approach
  • Detailed component breakdown
  • BDD specifications with Given-When-Then scenarios
  • Architecture patterns and integration points
  • Best practices and considerations
  • Error handling and edge cases
  • Testing strategy

• Design grounded in codebase reality

  • References specific files and patterns throughout
  • Shows concrete examples and interfaces
  • Aligns with project's architectural style

After Phase 4: Design Reflection

Before proceeding to Phase 5 (Git Commit), verify:

• Core reflection sub-agents launched and completed

  • Requirements Traceability Review sub-agent completed
  • BDD Completeness Review sub-agent completed
  • Cross-Document Consistency Review sub-agent completed

• Additional reflection sub-agents completed (if launched)

  • Security Review sub-agent completed (if applicable)
  • Risk Assessment sub-agent completed (if applicable)

• Findings synthesized and prioritized

  • All sub-agent outputs collected via TaskOutput
  • Findings merged into unified gap list
  • Gaps prioritized by impact (High/Medium/Low)

• Requirements traceability verified

  • Every Phase 1 requirement is addressed in design
  • Requirements are traced to specific sections
  • No orphaned requirements without implementation

• Gaps identified and filled

  • Error handling paths documented
  • Edge cases covered in BDD scenarios
  • Integration points clearly defined
  • Non-functional requirements addressed

• Cross-document consistency verified

  • Terminology consistent across all documents
  • Cross-references between documents work
  • Component and file names consistent

• BDD scenario completeness verified

  • Happy path scenarios covered
  • Error path scenarios covered
  • Edge case scenarios covered
  • Testing strategy is clear

• Documents updated based on findings

  • High priority gaps addressed
  • Gaps filled with new content
  • Ambiguities clarified
  • Inconsistencies fixed

After Phase 5: Git Commit

Before proceeding to Phase 6, verify:

• Entire folder committed

  • Used git add docs/plans/YYYY-MM-DD-<topic>-design/
  • NOT just individual files
  • Folder pattern matches requirements

• Commit message follows requirements

  • Prefix: docs: (lowercase)
  • Subject: Under 50 characters, lowercase
  • Body: User context, specific actions, design summary
  • Footer: Co-Authored-By with model name

• Commit verified

  • Ran git log -1 to confirm commit
  • Commit message displays correctly
  • All files included in commit

• User informed

  • Told user the folder and file location
  • Confirmed git commit completed
  • Ready to proceed with implementation planning

Success Indicators

High Quality Brainstorming Session:

• Explored codebase before asking questions
• Asked focused questions one at a time based on exploration gaps
• Proposed options grounded in existing patterns
• Got user buy-in before creating detailed design
• Launched core sub-agents in parallel for design creation
• Launched additional sub-agents as needed for specialized research
• Integrated all sub-agent results successfully
• Launched reflection sub-agents in parallel for gap identification
• Synthesized and prioritized reflection findings
• Updated design documents based on reflection findings
• Design document is comprehensive and actionable with BDD scenarios
• Saved design files with proper structure inside dated folder
• Committed design folder as checkpoint before implementation

Common Pitfalls to Avoid:

• Asking questions without exploring codebase first
• Asking multiple questions at once (always ask one at a time)
• Proposing abstract options not grounded in reality
• Creating design without user approval of approach
• Skipping alternatives without clear rationale
• Launching sub-agents without providing complete context
• Not integrating sub-agent results before creating documents
• Launching too many sub-agents for simple features (stay focused)
• Skipping reflection phase - "it looks fine, let's just commit"
• Single-agent reflection - not leveraging parallel sub-agents
• Ignoring sub-agent findings - not acting on identified gaps
• Superficial fixes - adding content without proper integration
• Missing BDD specifications or best practices sections
• Saving design to wrong location (should be docs/plans/YYYY-MM-DD-<topic>-design/)
• Committing individual files instead of entire folder
• Jumping to implementation without committing design