---

name: plan-execution description: Use when you have a written implementation plan to execute - supports both same-session subagent mode and separate-session batch mode

Plan Execution

Overview

Execute implementation plans task by task, with review between tasks.

Core principle: Fresh context per task + TDD + review gates = high quality output.

Default: TDD is mandatory unless user explicitly requests to skip it, or the task cannot be test-driven (e.g., config files, documentation, pure refactoring with existing tests).

Announce at start: "I'm using the plan-execution skill to implement this plan."

Choose Execution Mode

Subagent mode advantages: Fresh subagent per task, no context pollution, faster iteration Batch mode advantages: Human checkpoints between batches, better for complex/risky changes

---

Mode 1: Subagent Execution (Same Session)

Setup

1. Read plan file, extract ALL tasks with full text
2. Note context needed for each task
3. Create TodoWrite with all tasks

Per-Task Loop

4. Dispatch implementer subagent (use ./implementer-prompt.md)

5. Handle questions

• Implementer asks questions? → Answer clearly, provide context, re-dispatch
• No questions? → Continue

6. Implementer works (implements, tests, commits, self-reviews)

7. Spec compliance review

• Dispatch spec reviewer (use ./spec-reviewer-prompt.md)
• Reviewer approves? → Continue to step 8
• Issues found? → Implementer fixes → Re-review (repeat until ✅)

8. Code quality review

• Dispatch code quality reviewer (use ./code-quality-reviewer-prompt.md)
• Reviewer approves? → Mark task complete, go to step 9
• Issues found? → Implementer fixes → Re-review (repeat until ✅)

9. More tasks?

• Yes → Return to step 4
• No → Continue to Completion

Completion

1. Dispatch final code reviewer for entire implementation
2. Use harness:finishing-a-development-branch

---

Mode 2: Batch Execution (Separate Session)

Step 1: Load and Review Plan

1. Read plan file
2. Review critically - identify questions or concerns
3. If concerns: Raise with human partner before starting
4. If clear: Create TodoWrite and proceed

Step 2: Execute Batch (default: 3 tasks)

For each task:

1. Mark as in_progress
2. Follow each step exactly
3. Run verifications as specified
4. Mark as completed

Step 3: Report

When batch complete:

• Show what was implemented
• Show verification output
• Say: "Ready for feedback."

Step 4: Continue

Based on feedback:

• Apply changes if needed
• Execute next batch
• Repeat until complete

Step 5: Complete

Use harness:finishing-a-development-branch

---

When to Stop and Ask

STOP immediately when:

• Hit a blocker (missing dependency, test fails, instruction unclear)
• Plan has critical gaps
• You don't understand an instruction
• Verification fails repeatedly

Ask for clarification rather than guessing.

---

Red Flags

Never:

• Skip TDD without explicit user permission or valid exception
• Skip reviews (spec compliance OR code quality)
• Proceed with unfixed issues
• Dispatch multiple implementation subagents in parallel (conflicts)
• Make subagent read plan file (provide full text instead)
• Accept "close enough" on spec compliance
• Skip review loops when issues found
• Guess when blocked

If subagent asks questions: Answer clearly and completely before proceeding.

If reviewer finds issues: Implementer fixes → Reviewer re-reviews → Repeat until approved.

---

Prompt Templates

• ./implementer-prompt.md - Implementation subagent
• ./spec-reviewer-prompt.md - Spec compliance reviewer
• ./code-quality-reviewer-prompt.md - Code quality reviewer

---

Integration

Preceded by: harness:writing-plans Followed by: harness:finishing-a-development-branch Subagents MUST use: harness:test-driven-development (default, mandatory)