Claude Code Plugins

Community-maintained marketplace

Feedback

Human validation of PRD before design phase. Presents PRD, user approves or requests changes, extracts user stories. Required for complexity ≥15.

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 plan-approval
description Human validation of complete plan (PRD + Stories + Execution Plan) before execution. User sees full picture then approves or requests changes. Required for complexity ≥15.

Plan Approval Skill

ROOT AGENT ONLY - Uses AskUserQuestion, runs only from root agent.

Purpose: Get human approval of complete plan AFTER design phase Trigger: After design-phase completes (stories extracted, architect annotated, PM planned) Input: prdPath, storiesFolder, storyFiles, executionPlan, complexity Output: Approval status


Why Approval Happens Here (Not Earlier)

User gets to see the full picture before committing:

  1. ✅ PRD (requirements document)
  2. ✅ User stories (broken down by functional capability)
  3. ✅ Architecture annotations (on each story)
  4. ✅ Execution plan (waves, parallelization, dependencies)

This is more useful than approving just the PRD, because:

  • PRD alone doesn't show HOW implementation will happen
  • Stories show actual work breakdown and parallelization
  • Execution plan shows time investment and agent assignments
  • User can catch issues with story granularity before execution

When to Use

Complexity Use Plan Approval?
<15 No (skip)
≥15 Yes

Workflow

  1. Present complete plan to user:

    • Summary of PRD (link to file)
    • List of user stories with brief descriptions
    • Execution plan summary (waves, parallelization %)
    • Estimated agent count
  2. Ask for approval:

    PLAN SUMMARY
    ════════════════════════════════════════
    
    📄 PRD: docs/projects/msm008-feature/prd.md
    
    📋 User Stories (9 total):
    ├── Wave 1 (parallel): US-001, US-002
    ├── Wave 2 (parallel): US-003a, US-003b, US-003c, US-003d
    ├── Wave 3 (parallel): US-003e, US-004
    └── Wave 4: US-005
    
    ⚡ Parallelization: 65% (vs 0% if stories were per-package)
    🤖 Max concurrent agents: 4
    
    Ready to execute?
    
    • YES → Continue to execution-phase
    • NO → Ask what changes needed → Go to step 3
  3. If changes requested:

    • Identify which component needs revision (PRD, stories, or plan)
    • Spawn appropriate agent (BA for PRD/stories, PM for plan)
    • Update files
    • Loop back to step 1 (max 2 iterations)
  4. Continue to execution-phase with approved plan


Output Format

{
  "status": "approved",
  "prdPath": "docs/projects/msm008-feature/prd.md",
  "storiesFolder": "docs/projects/msm008-feature/user-stories/",
  "storyFiles": [
    "US-001-database-schema.md",
    "US-002-contracts-types.md",
    "US-003a-workflow-scaffolding.md",
    "US-003b-height-weight-parser.md",
    "US-003c-team-fuzzy-matching.md",
    "US-003d-major-entity-parser.md",
    "US-003e-validation-upsert.md",
    "US-004-api-layer.md",
    "US-005-frontend.md"
  ],
  "executionPlan": "docs/projects/msm008-feature/execution-plan.md",
  "iterations": 1
}

Revision Scenarios

User Feedback Action
"Stories too coarse" BA re-extracts with finer granularity
"Wrong architecture approach" Architect revises annotations
"Too many waves, can we parallelize more?" PM revises execution plan
"Missing acceptance criteria" BA updates specific story files
"PRD scope changed" BA updates PRD, re-extract stories, re-plan

Integration

Called by: /audit, /build, /ms (complexity ≥15) Calls: business-analyst agent (revise), architect agent (revise), project-manager agent (revise), AskUserQuestion Previous phase: design-phase (stories extracted, annotated, planned) Next phase: execution-phase