Claude Code Plugins

Community-maintained marketplace

Feedback

Intelligently navigate the 7-phase Spec Kit → CCGC workflow by detecting current phase, validating prerequisites, recommending transitions, and tracking per-project state. Use when the user asks "what phase", "what's next", "workflow steps", "guide me through pipeline", or phase keywords like "specify", "plan", "tasks", "implement", "validate", "sync". Essential for multi-project developers maintaining context across different features and codebases simultaneously.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name Pipeline Phase Guide
description Intelligently navigate the 7-phase Spec Kit → CCGC workflow by detecting current phase, validating prerequisites, recommending transitions, and tracking per-project state. Use when the user asks "what phase", "what's next", "workflow steps", "guide me through pipeline", or phase keywords like "specify", "plan", "tasks", "implement", "validate", "sync". Essential for multi-project developers maintaining context across different features and codebases simultaneously.
allowed-tools Read, Write, Glob, Bash

Pipeline Phase Guide

Purpose

This Skill provides intelligent navigation through your 7-phase Spec Kit + CCGC pipeline:

Phase 0 → Phase 1 → Phase 2 → Phase 3 → Phase 4 → Phase 5 → Phase 6 → (repeat)
        PLANNING PHASES              IMPLEMENTATION PHASES
   └─────────────────────────────────────────────────────┘
          Spec Kit (what/how)        CCGC (build/validate)

Key Problem Solved: Maintains context across multiple projects with different phases, preventing confusion about "which feature is at which phase?"


The 7-Phase Pipeline Explained

Phase 0: Undefined (No Feature Active)

  • When: Starting new feature or completed previous feature
  • Files: None yet
  • Next: Create specification
  • Slash Command: /specify "feature description"
  • Time: Variable

Phase 1: Specification (What to Build?)

  • When: User has written feature specification
  • Files: specs/###-feature-name/spec.md exists
  • Output: Feature ID, scope, requirements, success criteria
  • Next: Design the implementation
  • Slash Command: /plan
  • Time: 1-2 hours

Phase 2: Planning (How to Design?)

  • When: Design decisions made, architecture defined
  • Files: plan.md created, auto-synced to memory/architecture.md
  • Output: Technical approach, file structure, data model, API contracts
  • Next: Break down into tasks
  • Slash Command: /tasks
  • Time: 2-4 hours

Phase 3: Task Breakdown (What's the Work?)

  • When: Tasks generated and ready for implementation
  • Files: tasks.md created, auto-synced to memory/tasks_plan.md
  • Output: TDD-ordered tasks, dependencies, time estimates
  • Next: Start implementation
  • Slash Command: @agent-implementer or say "ready to code"
  • Time: 30 minutes

Phase 4: Implementation (Build It!)

  • When: Tasks being executed, code being written
  • Files: Source code files modified, tests created
  • Output: New feature, tests passing
  • Next: Run validation suite
  • Slash Command: /test-run, /standards-check
  • Time: 4-8 hours (depends on feature)

Phase 5: Validation (Did We Get It Right?)

  • When: Code complete, tests passing, standards checked
  • Files: All tests passing, no PHPStan/ESLint errors
  • Output: Validated implementation, ready for deployment
  • Next: Capture learnings
  • Slash Command: /memory-sync
  • Time: 1-2 hours

Phase 6: Learning & Completion (What Did We Learn?)

  • When: Memory synced, feature complete
  • Files: lessons-learned.md updated, error-documentation.md updated
  • Output: Team insights, documented patterns
  • Next: Create PR or start new feature
  • Action: Merge to main or create PR
  • Time: 30 minutes

How It Works

Automatic Phase Detection

The Skill determines current phase by checking file existence:

Check sequence:
1. Does specs/###-feature-name/spec.md exist?
   NO  → Phase 0 (Undefined)
   YES → Continue

2. Does plan.md exist in spec directory?
   NO  → Phase 1 (Specification complete, planning needed)
   YES → Continue

3. Does tasks.md exist?
   NO  → Phase 2 (Planning complete, tasks needed)
   YES → Continue

4. Are there incomplete tasks in tasks_plan.md?
   YES → Phase 4 (Implementation in progress)
   NO  → Check tests

5. Are tests passing?
   NO  → Phase 4 (Still implementing)
   YES → Phase 5 (Validation phase)

6. Is memory synced?
   NO  → Phase 5 (Validation, need sync)
   YES → Phase 6 (Complete)

Phase-Specific Guidance

For each phase, the Skill provides:

  1. Current State: What's been completed
  2. Next Action: Specific next step (slash command or user action)
  3. Prerequisites: What must be done first
  4. Estimated Time: How long this phase typically takes
  5. Warnings: Any blockers or issues

Transition Validation (Enhanced v2.0.0)

Before recommending transition to next phase, Skill validates:

Phase 0 → 1 (Start Specification):

  • Git repo clean
  • No conflicting features in progress
  • Constitution ratified (optional, NEW v2.0.0)

Phase 1 → 2 (Start Planning):

  • Spec complete (no [NEEDS CLARIFICATION] markers)
  • Stakeholder approval obtained
  • Requirements clear and achievable
  • Quality Gate 1 - Clarity Gate (NEW v2.0.0):
    • Requirements unambiguous
    • Success criteria measurable
    • Scope well-defined

Phase 2 → 3 (Start Task Breakdown):

  • Plan has Constitution Check completed (if constitution exists)
  • Technical approach finalized
  • No outstanding questions
  • Quality Gate 2 - Testability Gate (NEW v2.0.0):
    • Test strategy defined
    • Test framework selected
    • TDD approach documented
  • Quality Gate 3 - Boundary Gate (NEW v2.0.0):
    • Architectural boundaries defined
    • Component interfaces specified
    • No circular dependencies

Phase 3 → 4 (Start Implementation):

  • Memory context loaded (/memory-read run)
  • No conflicting changes on main branch
  • Development environment ready
  • Tests written (TDD requirement, NEW v2.0.0)
  • Plugins loaded (if project uses plugins, NEW v2.0.0)
  • Validators warmed (if plugins exist, NEW v2.0.0)

Phase 4 → 5 (Start Validation):

  • All tasks completed
  • All tests passing
  • Code standards passing (linter/formatter)
  • No uncommitted changes
  • Quality Gate 4 - Coverage Gate (NEW v2.0.0):
    • Test coverage >80%
    • Critical paths covered
    • Edge cases tested
  • Quality Gate 5 - Security Gate (NEW v2.0.0):
    • No SQL injection risks
    • No hardcoded secrets
    • Input validation implemented

Phase 5 → 6 (Start Learning):

  • Tests passing (confirmed)
  • Performance targets met
  • Documentation complete
  • Ready for memory sync
  • Quality Gate 6 - Maintainability Gate (NEW v2.0.0):
    • Code reviewed (self or peer)
    • Complex logic documented
    • Technical debt minimized

Phase 6 → 0 (Complete Feature):

  • PR merged or changes committed
  • Memory synced
  • Lessons documented
  • Quality Gate 7 - Documentation Gate (NEW v2.0.0):
    • Lessons learned captured
    • Patterns documented
    • Known issues recorded

Examples

Example 1: Initial Feature Discovery

User Input: "What phases do I need to go through?"

Skill Output:

🧭 SPEC KIT + CCGC PIPELINE (7 Phases)
════════════════════════════════════════════════════════════════

┌─ PLANNING (Spec Kit - Upstream) ─┐
│                                   │
│  Phase 1: SPECIFICATION           │
│  └─ Create spec.md with feature   │
│     requirements, scope, success  │
│     criteria                       │
│                                   │
│  Phase 2: PLANNING                │
│  └─ Design implementation         │
│     approach, file structure,     │
│     data model, API contracts     │
│                                   │
│  Phase 3: TASK BREAKDOWN          │
│  └─ Break design into tasks       │
│     (TDD-ordered, estimated)      │
│                                   │
└─ IMPLEMENTATION CHECKPOINT ────────┘
              ↓
┌─ IMPLEMENTATION (CCGC - Downstream) ┐
│                                      │
│  Phase 4: IMPLEMENTATION             │
│  └─ Execute tasks, write code,       │
│     tests before implementation      │
│                                      │
│  Phase 5: VALIDATION                 │
│  └─ Run tests, standards checks,     │
│     verify performance               │
│                                      │
│  Phase 6: LEARNING                   │
│  └─ Capture insights,                │
│     documented patterns,             │
│     sync memory                      │
│                                      │
└─ COMPLETION CHECKPOINT ──────────────┘
              ↓
      Start next feature (Phase 0)

🎯 EACH PHASE INCLUDES:
   - Specific slash commands to execute
   - Files created as outputs
   - Quality gates to pass
   - Estimated time

💡 QUICK START:
   New feature? → /specify "description"
   Have spec? → /plan
   Have plan? → /tasks
   Ready to code? → "ready to code" (quick-start Skill)
════════════════════════════════════════════════════════════════

Example 2: Multi-Project Phase Tracking

User Input: "What phases are my projects in?"

Skill Output:

🧭 MULTI-PROJECT PHASE STATUS
════════════════════════════════════════════════════════════════

Project 1: /Users/dev/dev_IECPN
├─ Current Phase: 4 - IMPLEMENTATION (Active)
├─ Feature: #006 - PHI Logging Cleanup
├─ Progress: 0/26 tasks complete
├─ Next: @agent-implementer
└─ Status: Ready to start

Project 2: /Users/dev/api-refactor
├─ Current Phase: 2 - PLANNING (Active)
├─ Feature: #012 - API Gateway
├─ Progress: plan.md created, no tasks
├─ Next: /tasks to generate breakdown
└─ Status: Planning phase complete

Project 3: /Users/dev/ml-pipeline
├─ Current Phase: 1 - SPECIFICATION (Paused)
├─ Feature: #008 - ML Model Pipeline
├─ Progress: spec.md created
├─ Next: /plan to design approach
└─ Status: Waiting for planning

💡 NEXT ACTIONS BY PROJECT:
   dev_IECPN → Start: @agent-implementer (immediate)
   api-refactor → Run: /tasks (before next work)
   ml-pipeline → Continue: /plan (when ready)
════════════════════════════════════════════════════════════════

Example 3: Guided Phase Transition with Quality Gates (NEW v2.0.0)

User Input: "I'm ready to move from planning to implementation"

Skill Output:

🧭 PHASE TRANSITION: Planning → Implementation (v2.0.0)
════════════════════════════════════════════════════════════════

Current: Phase 2 - PLANNING
Target: Phase 3-4 - TASK BREAKDOWN → IMPLEMENTATION

✅ Prerequisites Validation:
   [✓] Plan created (plan.md exists)
   [✓] Technical approach finalized
   [✓] No outstanding questions
   [✓] Git working tree clean

⚖️  QUALITY GATES VALIDATION (NEW v2.0.0):
   Constitution: Found (7 principles, 7 gates)

   Gate 2 - Testability Gate:
   [✓] Test strategy defined (TDD with pytest)
   [✓] Test framework documented in plan.md
   [✓] Test-first approach specified in tasks
   Status: ✅ PASSED

   Gate 3 - Boundary Gate:
   [✓] Architectural boundaries defined
   [✓] Component interfaces specified in architecture.md
   [✓] No circular dependencies detected
   Status: ✅ PASSED

   Overall: 2/2 gates passed for this transition
   Status: ✅ CLEAR TO PROCEED

🎯 REQUIRED NEXT STEP:
   Generate task breakdown: /tasks

   This will:
   1. Parse plan.md for implementation approach
   2. Generate TDD-ordered task list
   3. Estimate effort per task
   4. Validate against quality gates
   5. Create tasks.md in specs/###-feature/
   6. Auto-sync to memory/tasks_plan.md

⏱️  Estimated: 5-10 minutes

🎯 AFTER /tasks COMPLETES:
   Say: "ready to code"
   This will activate quick-start Skill to:
   1. Load full development context
   2. Validate phase and prerequisites
   3. Run constitutional pre-flight checks
   4. Load project plugins (if any)
   5. Warm validator cache
   6. Present next task
   7. Confirm readiness
   8. Guide to @agent-implementer

📊 TYPICAL FLOW (v2.0.0):
   /plan → (create plan.md)
   /constitution-check → (validate against gates) [if constitution exists]
   /tasks → (create tasks.md with quality gate validation)
   "ready to code" → (quick-start with plugin loading)
   @agent-implementer → (execute tasks)

Ready to proceed?

With Quality Gate Failure:

🧭 PHASE TRANSITION: Planning → Implementation
════════════════════════════════════════════════════════════════

❌ TRANSITION BLOCKED

⚖️  QUALITY GATES VALIDATION:
   Gate 2 - Testability Gate:
   [✓] Test strategy defined
   [❌] Test framework NOT documented
   [❌] TDD approach missing from plan
   Status: ❌ FAILED

   Gate 3 - Boundary Gate:
   [✓] Architectural boundaries defined
   [⚠️] Component interfaces partially specified
   [✓] No circular dependencies
   Status: ⚠️ WARNING

   Overall: 0/2 gates passed, 1 warning
   Status: ❌ CANNOT PROCEED

🎯 REQUIRED ACTIONS:
   1. Update plan.md with:
      - Test framework selection (pytest/jest/etc)
      - TDD approach documentation
      - Test-first task ordering

   2. Fix warnings:
      - Complete component interface specifications
      - Document API contracts in architecture.md

   3. Re-run: "ready to move from planning to implementation"
      to re-validate quality gates

⏱️  Estimated fix time: 15-20 minutes

📚 REFERENCE:
   Constitution: .specify/memory/constitution.md
   Principles: II (Testability), III (Boundary Respect)
   Quality Gate Docs: Check principle requirements

Cannot proceed until gates pass.

Example 4: Phase Mismatch Detection

User Input: "Start coding" (but actually in planning phase)

Skill Output:

⚠️  PHASE MISMATCH

Current State:
├─ Phase: 2 - PLANNING
├─ Feature: #012 - API Gateway
└─ Issue: Can't implement without task breakdown

What's Been Done:
✓ Specification (spec.md)
✓ Planning (plan.md)

What's Missing:
❌ Task Breakdown (tasks.md)

🎯 REQUIRED STEP:
   Run: /tasks

   This generates:
   - TDD-ordered task list
   - Effort estimates
   - Dependency information
   - Parallel opportunities

⏱️  Time: ~5 minutes

After /tasks:
   Then: "ready to code" → Implementation

Can't skip this step because:
   - @agent-implementer requires tasks.md
   - No tasks = no execution plan
   - Task breakdown is critical quality gate

Proceed?

Cross-Project State Management

State Tracking File

Location: ~/.claude/skills/phase-guide/state.json

Purpose: Remember phase for each project

Format:

{
  "projects": {
    "/Users/dev/dev_IECPN": {
      "currentPhase": 4,
      "featureId": "006",
      "featureName": "PHI Logging Cleanup",
      "branch": "006-feature-006-logging",
      "lastUpdate": "2025-10-17T10:30:00Z",
      "contextFreshness": "current",
      "lastAction": "@agent-implementer started"
    },
    "/Users/dev/api-refactor": {
      "currentPhase": 2,
      "featureId": "012",
      "featureName": "API Gateway",
      "branch": "012-api-gateway",
      "lastUpdate": "2025-10-16T14:15:00Z",
      "contextFreshness": "stale",
      "lastAction": "/plan completed"
    }
  },
  "lastActiveProject": "/Users/dev/dev_IECPN",
  "lastSync": "2025-10-17T10:30:00Z"
}

Auto-Update Mechanism

State file updates when:

  1. Phase transition detected
  2. New feature started
  3. User switches projects
  4. Memory sync occurs
  5. Major milestone reached

Multi-Project Context Switching

Scenario: User switches from Project A to Project B

Skill Behavior:

  1. Detects new project (different directory)
  2. Looks up Project B state in state.json
  3. Reports previous phase for Project B
  4. Shows tasks left off at
  5. Guides resumption

Integration with Other Skills

Relationship to context-navigator:

  • context-navigator: "Where am I?" (current snapshot)
  • phase-guide: "What's next?" (workflow guidance)
  • Flow: See state → Understand phase → Navigate transitions

Relationship to quick-start:

  • phase-guide: Detects if you're in right phase for coding
  • quick-start: Validates and loads context
  • Flow: phase-guide says "ready", quick-start confirms "READY"

Relationship to memory-keeper:

  • phase-guide: Tracks phase transitions
  • memory-keeper: Keeps memory fresh during phase
  • Flow: Transition occurs → memory-keeper keeps data current

Troubleshooting

Issue: Skill says you're in Phase 1 but you know tasks exist

Cause: tasks.md might be in wrong location

Check:

find . -name "tasks.md" -type f
# Should find: specs/###-feature-name/tasks.md

Fix: Verify tasks.md is in correct location


Issue: State.json shows wrong phase for a project

Cause: Phase detection failed or state not synced

Fix: Say "sync state" to force re-detect and update


Performance

This Skill performs:

  • Phase detection (file checks)
  • State file reads/writes
  • Cross-project tracking
  • Quality gate validation (NEW v2.0.0)
  • Constitution parsing (NEW v2.0.0)

Expected execution time: <300ms (with constitution and quality gates) Expected execution time: <200ms (v1.0.0 mode, no constitution) Side effects: Updates state.json (intentional)


Graceful Degradation (NEW v2.0.0)

This skill is backward compatible and works without v2.0.0 features:

Without Constitution (v1.0.0 mode):

⚖️  QUALITY GATES VALIDATION
   Constitution: Not ratified
   Quality Gates: Skipped (v1.0.0 compatibility)
   Status: ✅ Proceeding without constitutional governance
  • All phase transitions work normally
  • Prerequisites checked as in v1.0.0
  • No quality gate validation

Without Quality Gate Cache:

  • Reads constitution directly (slower by ~50ms)
  • Parses principles on-demand
  • No functional differences

Partial Quality Gate Failures:

  • Shows which gates passed/failed
  • Provides specific remediation steps
  • Blocks transition only if CRITICAL gates fail
  • WARN/MEDIUM gates show warnings but allow proceed

Missing v2.0.0 Directories:

No .claude/cache/principle-registry.json → Parse constitution directly
No .specify/memory/constitution.md → Skip quality gates, v1.0.0 mode

Multi-Project State Handling:

  • Works with or without state.json
  • Falls back to file-based detection if state missing
  • Creates state.json on first use

This ensures:

  1. Smooth v1.0.0 → v2.0.0 migration
  2. Works without constitution
  3. Quality gates optional, not required
  4. No breaking changes

What This Skill Does NOT Do

  • ❌ Does NOT execute tasks (use @agent-implementer)
  • ❌ Does NOT run /memory-sync (use memory-keeper Skill or /memory-sync)
  • ❌ Does NOT modify code (read-only diagnostic)
  • ❌ Does NOT make decisions for you (guidance only)
  • ❌ Does NOT enforce quality gates (only validates and warns) (NEW v2.0.0)
  • ❌ Does NOT fix constitutional violations (provides remediation guidance)

This Skill navigates workflow. Slash commands and agents execute the work.