SpecsImplementationScope ControlAI Agent Skill
spec-driven-implementation
Turns a finished spec into scoped implementation steps and flags any request that falls outside what the spec actually asked for.
Install Options
Install into Claude Code, Cursor, Codex, or Antigravity with:
$ npx skills add warpdotdev/common-skills@spec-driven-implementation
Real Work Situation Solved
Turn a vague feature ask into a clear spec
“Your lead says "make the product smarter with AI" - where do you start?”
5-Step Action Framework
- 1Rewrite the request as one measurable goal statement - no vague adjectives like "smarter" or "better".
- 2List concrete input/output: what data goes in, what result comes out, who the end user is.
- 3Mark the boundary between the happy path and the error/edge cases that need handling.
- 4Define a "done" criterion both you and engineering can agree on.
- 5Send the draft to one engineer before the meeting - if they come back with more than 3 questions, the spec isn't clear enough yet.
Before
"Make search smarter with AI."
After
"After the user types 3+ characters, return up to 5 suggestions ranked by relevance in under 300ms; if there are no matches, show trending queries instead of a blank screen."
Related Course Lesson
AI Agents for PM • Lesson 17
SKILL.md Source Instructions
SKILL.md • Read-only preview# Spec-Driven Implementation Skill Read the spec fully before writing code, break it into scoped tasks, and flag any request that falls outside the spec's stated boundary instead of silently expanding scope.