Load plan, review critically, execute all tasks, report when complete.
Announce at start: "I'm using the executing-plans skill to implement this plan."
Prefer subagent-driven-development over this skill when subagents are available —
it provides higher quality through per-task fresh context and two-stage review.
The Process
Step 1: Load and Review Plan
- Read plan file completely
- Review critically — identify questions or concerns about the plan
- If concerns: raise with human partner BEFORE starting
- If no concerns: create TodoWrite and proceed
Step 2: Execute Tasks
For each task:
- Mark as in_progress
- Follow each step exactly (plan has bite-sized steps)
- Run verifications as specified
- Mark as completed
Do not pause to check in between tasks unless blocked.
Step 3: Complete Development
After all tasks complete:
- Use
finishing-a-development-branchskill - Verify tests, update docs, commit
When to Stop and Ask
STOP executing immediately when:
- Hit a blocker (missing dependency, test fails, instruction unclear)
- Plan has critical gaps preventing starting
- Verification fails repeatedly
Ask for clarification rather than guessing.
Hard Rules
- Never start on main/master branch without explicit user consent
- Follow plan steps exactly — don't improvise
- Stop when blocked, don't guess
- Skip nothing in the verification steps