Agent skill
Phase Management
This skill should be used when the user asks about "phase management", "6-phase protocol", "explore phase", "plan phase", "code phase", "test phase", "fix phase", "document phase", "phase transitions", "phase enforcement", or needs guidance on managing the mandatory orchestration phases.
Install this agent skill to your Project
npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/data/phase-management
SKILL.md
Phase Management
Establish systematic progression through the mandatory 6-phase protocol for all complex tasks, ensuring quality gates and proper transitions.
The 6-Phase Protocol
Every complex task MUST progress through these phases in order:
EXPLORE → PLAN → CODE → TEST → FIX → DOCUMENT
Phase Overview
| Phase | Purpose | Min Agents | Outputs |
|---|---|---|---|
| EXPLORE | Understanding & Research | 2+ | Context, requirements, constraints |
| PLAN | Strategy & Design | 1-2 | Architecture, task breakdown, DAG |
| CODE | Implementation | 2-4 | Working code, initial tests |
| TEST | Validation | 2-3 | Test results, coverage reports |
| FIX | Refinement | 1-2 | Bug fixes, improvements |
| DOCUMENT | Knowledge Transfer | 1-2 | Docs, ADRs, vault updates |
Phase 1: EXPLORE
Purpose: Gather comprehensive understanding before planning.
Activities
- Analyze requirements and constraints
- Research existing codebase
- Identify dependencies and risks
- Gather library documentation (Context7)
- Understand domain context
Agents
code-explorer- Codebase analysisresearch-agent- Documentation gatheringrequirements-analyst- Requirement extraction
Exit Criteria
- Requirements documented
- Codebase context understood
- Risks identified
- Dependencies mapped
Outputs
explore_output:
requirements: "Detailed requirement specification"
codebase_context: "Relevant files and patterns"
risks: "Identified risks and mitigations"
dependencies: "External and internal dependencies"
Phase 2: PLAN
Purpose: Design solution and create execution plan.
Activities
- Design architecture/approach
- Break down into tasks
- Create dependency graph (DAG)
- Allocate resources
- Estimate complexity
Agents
master-strategist- Overall planningarchitect-supreme- Technical design
Exit Criteria
- Architecture defined
- Tasks broken down
- DAG created
- Resources allocated
Outputs
plan_output:
architecture: "Technical design document"
task_dag: "Directed acyclic graph of tasks"
resource_allocation: "Agent assignments"
checkpoints: "Phase boundary markers"
Phase 3: CODE
Purpose: Implement the planned solution.
Activities
- Execute tasks per DAG
- Write production code
- Write unit tests
- Handle edge cases
- Integrate components
Agents
coder- Implementation (2-4 agents)unit-tester- Test writingintegrator- Component integration
Exit Criteria
- All tasks completed
- Unit tests passing
- Code compiles/runs
- Integration successful
Outputs
code_output:
files_created: ["list of new files"]
files_modified: ["list of changed files"]
tests_written: ["list of test files"]
integration_status: "success|partial|failed"
Phase 4: TEST
Purpose: Validate implementation quality.
Activities
- Run full test suite
- Perform integration testing
- Execute E2E tests (if applicable)
- Measure coverage
- Security scanning
Agents
test-runner- Test executioncoverage-analyst- Coverage analysissecurity-scanner- Security validation
Exit Criteria
- All tests pass
- Coverage meets threshold
- No security issues
- Performance acceptable
Outputs
test_output:
test_results: "pass|fail with details"
coverage_report: "Coverage percentage by file"
security_findings: "Any vulnerabilities found"
performance_metrics: "Key performance indicators"
Phase 5: FIX
Purpose: Address issues found during testing.
Activities
- Fix failing tests
- Address security findings
- Improve coverage
- Optimize performance
- Refactor as needed
Agents
debugger- Issue diagnosisfixer- Bug resolution
Exit Criteria
- All tests passing
- Security issues resolved
- Coverage improved
- Performance optimized
Outputs
fix_output:
issues_resolved: ["list of fixed issues"]
regressions_avoided: "Verification status"
final_test_status: "All tests passing confirmation"
Phase 6: DOCUMENT
Purpose: Create knowledge transfer artifacts.
Activities
- Update README/docs
- Create ADRs for decisions
- Sync to Obsidian vault
- Update API documentation
- Log learnings
Agents
documentation-expert- Doc writingvault-syncer- Obsidian sync
Exit Criteria
- Docs updated
- ADRs created
- Vault synced
- Knowledge captured
Outputs
document_output:
docs_updated: ["list of documentation files"]
adrs_created: ["architectural decision records"]
vault_entries: ["Obsidian vault updates"]
commit_ready: "Ready for final commit"
Phase Transitions
Transition Protocol
- Verify Exit Criteria: All criteria must be met
- Create Checkpoint: Save phase state
- Package Outputs: Prepare for next phase
- Notify Transition: Log phase completion
- Initialize Next Phase: Load context, spawn agents
Checkpoint Structure
checkpoint:
phase: "EXPLORE|PLAN|CODE|TEST|FIX|DOCUMENT"
timestamp: "ISO 8601 timestamp"
status: "completed|partial|blocked"
outputs: "Phase output package"
next_phase: "Next phase to execute"
context: "Preserved context for continuation"
Recovery from Checkpoint
If interrupted, resume from last checkpoint:
- Load checkpoint state
- Verify outputs still valid
- Continue from current phase
- Don't repeat completed work
Enforcement
The 6-phase protocol is MANDATORY and enforced via hooks:
- Pre-task: Verify phase compliance plan
- Post-task: Validate all phases completed
- Phase-transition: Enforce exit criteria
Skip Prevention
Phases cannot be skipped:
- TEST phase: Required even for "simple" changes
- DOCUMENT phase: Required for all work
- FIX phase: Can be minimal if no issues
Additional Resources
Reference Files
references/phase-details.md- Extended phase specificationsreferences/checkpoint-format.md- Checkpoint data structures
Examples
examples/phase-outputs.json- Sample outputs by phaseexamples/checkpoint.json- Sample checkpoint structure
Didn't find tool you were looking for?