blueprint-prp-execute

/blueprint:prp-execute

Execute a PRP (Product Requirement Prompt) with systematic implementation, validation gates, TDD workflow, and quality assurance.

Usage: /blueprint:prp-execute [prp-name]

Prerequisites:

• PRP exists in docs/prps/[prp-name].md
• Confidence score >= 7 (if lower, suggest /blueprint:prp-create refinement)

For detailed report templates, deferred items workflow, feature tracker sync, and error handling patterns, see REFERENCE.md.

When to Use This Skill

Use this skill when...	Use alternative when...
Ready to implement a planned feature from a PRP	PRP is not yet ready (confidence < 7)
Want to execute with full validation and TDD workflow	Implementing ad-hoc features without documentation
Need feature tracker and GitHub issue tracking	Working on isolated bug fixes
Want automatic progress reporting and deferred items tracking	Quick prototyping without formal tracking

Context

• PRP file path: !find . -maxdepth 3 -path "./docs/prps/${1:-unknown}.md"
• PRP confidence score: !find . -path '*/docs/prps/*' -maxdepth 3 -name "${1:-unknown}.md" -exec grep -m1 "^confidence:" {} +
• Feature tracker enabled: !find . -path '*/docs/blueprint/*' -maxdepth 3 -name 'feature-tracker.json' -type f
• Current branch: !git rev-parse --abbrev-ref HEAD
• Uncommitted changes: !git status --porcelain

Parameters

Parse $ARGUMENTS:

• prp-name (required): Name of PRP file in docs/prps/ (without .md extension)
  • Example: feature-auth-oauth2 → loads docs/prps/feature-auth-oauth2.md

Execution

Execute the complete PRP implementation workflow:

Step 1: Load and validate PRP

1. Read PRP file: docs/prps/{prp-name}.md
2. Extract confidence score from frontmatter
3. If confidence < 7 → Error: "PRP confidence too low. Run /blueprint:prp-create {prp-name} to refine"
4. Check content quality with a fresh reader (issue #3002). The confidence score and the structural hook (check-prp-readiness.sh) do not judge whether the content can be implemented, and the agent about to execute the PRP has a stake in calling it ready (.claude/rules/loop-integrity.md Pillar 1). Dispatch one Task agent (subagent_type: general-purpose) and give it only the PRP file path and the reader prompt in REFERENCE.md: no session context, no plan. It judges specific requirements, testable acceptance criteria, edge cases, and bounded scope, and returns a JSON verdict.
   • {"ok": true} → continue
   • {"ok": false, "reason": ...} → stop before any execution or delegation, print the reason, and suggest /blueprint:prp-create {prp-name} to refine the PRP
5. If confidence >= 9 → Offer delegation: "This PRP has high confidence. Execute now (current session) or create work-order for delegation?"
   • If work-order chosen → Run /blueprint:work-order --from-prp {prp-name} and exit
   • If delegation to multiple subagents chosen → Create focused work-orders per module from Implementation Blueprint and exit
6. Continue to Step 2 if executing now OR confidence 7-8
7. Load all referenced curated rule entries (.claude/rules/) for context
8. Parse the Implementation Blueprint into an ordered task checklist (dependencies first); track it with TodoWrite when the session has the task tools, otherwise carry it as a checklist in the response (see .claude/rules/agentic-permissions.md § Task-tool availability)

Step 2: Establish baseline with validation gates

Run pre-implementation validation gates (see REFERENCE.md) to establish clean starting state:

1. Linting gate: [command from PRP] - Expected: PASS
2. Existing tests gate: [command from PRP] - Expected: PASS

If gates fail:

• Document existing issues
• Decide: fix first or proceed with note
• Continue when ready

Step 3: Execute TDD implementation cycle

For each task in Implementation Blueprint:

1. RED phase: Write failing test matching PRP TDD Requirements

   • Create test file if needed
   • Run tests → Confirm FAILURE (test is meaningful)

2. GREEN phase: Implement minimal code to pass test

   • Follow Codebase Intelligence patterns from PRP
   • Apply patterns from curated rule references
   • Watch for Known Gotchas
   • Run tests → Confirm SUCCESS

3. REFACTOR phase: Improve code while keeping tests green

   • Extract common patterns
   • Improve naming, add type hints
   • Follow project conventions
   • Stay inside the task's files and the PRP task list. A pre-existing bug or unrelated cleanup you notice is not part of this change — record it as a deferred item in Step 5 (a GitHub issue if it is Blocked or Phase 2, otherwise Nice-to-Have). Do not commit scratch or exploratory tests.
   • Run tests → Confirm PASS
   • Run validation gates frequently (not just at end)

4. Mark progress: once the task's tests pass, tick Task N in the checklist from Step 1 (via TodoWrite when the task tools are present).

Step 4: Run comprehensive final validation

Execute all validation gates from PRP (see REFERENCE.md):

• Linting: [cmd] - Expected: PASS
• Type checking: [cmd] - Expected: PASS
• Unit tests: [cmd] - Expected: PASS (all tests)
• Integration tests: [cmd] - Expected: PASS (if applicable)
• Coverage check: [cmd] - Expected: Meets threshold
• Security scan: [cmd] - Expected: No high/critical issues (if applicable)
• Performance tests: [cmd] - Expected: Meets baseline (if defined)

Verify each success criterion from PRP.

Step 5: Document deferred items

Identify and track any deferred work:

1. Review Implementation Blueprint - items not completed
2. Categorize each deferred item:
   • Phase 2 (Required): Must have GitHub issues created
   • Nice-to-Have: Optional, no issue required
   • Blocked: Cannot complete - document blocker, create issue
3. Create GitHub issues for all Phase 2 and Blocked items (see REFERENCE.md)
4. Update PRP with deferred items section linking to GitHub issues
5. Do NOT proceed to Step 6 until all required issues are created

Step 6: Sync feature tracker (if enabled)

If feature tracker exists (docs/blueprint/feature-tracker.json):

1. Identify which feature codes (e.g., FR2.1) were addressed from PRP
2. Update feature tracker for each code:
   • Status: complete (all criteria met) or partial (some criteria met) or in_progress
   • Files: List of modified/created files
   • Tests: List of test files
   • Commits: Commit hashes
   • Notes: Implementation notes
3. Recalculate statistics: completion percentages, phase status
4. Update TODO.md: Check boxes for completed features
5. Report feature tracker changes

Step 7: Report results and next steps

Generate comprehensive execution summary report:

• Tasks completed: X/Y
• Tests added: N
• Files modified: [list]
• Validation results: Table of all gates (PASS/FAIL status)
• Success criteria: All verified
• Deferred items summary: Count and GitHub issue numbers
• Feature tracker updates: Features updated and percentages
• New gotchas discovered: [documented for future reference]
• Recommendations: Follow-up work or curated rule updates

Prompt user for next action:

• Commit changes (Recommended) → Run /git:commit
• Create work-order for follow-up → Run /blueprint:work-order
• Capture patterns as curated rules → Run /blueprint:curate-docs
• Continue to next PRP → Run /blueprint:prp-execute [next]
• Done for now → Exit

Agentic Optimizations

Context	Command
Check PRP exists	test -f docs/prps/${1}.md && echo "EXISTS" || echo "MISSING"
Extract confidence	head -50 docs/prps/${1}.md | grep -m1 "^confidence:" | sed 's/^[^:]*:[[:space:]]*//'
List all PRPs	ls docs/prps/*.md 2>/dev/null | xargs basename -s .md
Check feature tracker	test -f docs/blueprint/feature-tracker.json && echo "YES" || echo "NO"
Fast validation	Run validation gates in parallel when possible

---

For detailed validation gate definitions, deferred items workflow, error handling procedures, and agent team coordination, see REFERENCE.md.