Write plans a context-free agent could execute

Steps

1. 1Split work into tasks completable in 2-5 minutes by a fresh agent
2. 2Each task lists exact file paths and complete expected content — zero-context assumption
3. 3Attach a verification step per task: command + expected observable output
4. 4Order tasks so every dependency precedes its consumer; flag exceptions loudly
5. 5Include the 'why' once at plan top; tasks carry only the 'what'
6. 6Dry-run read: could an eager junior with no project context follow task #7 alone?

Common Pitfalls

• ▲Tasks referencing 'as discussed' knowledge agents don't have
• ▲Verification steps like 'check it works' proving nothing

Commands

$ npx skills add aniruddhaadak80/skills --skill design-gates-implementation-plan-writing

$ npx skills add aniruddhaadak80/skills --skill design-gates-implementation-plan-writing -g

Tags