Spec

When you're about to build something non-trivial and want a human checkpoint before the AI starts writing code, this skill generates a structured technical spec in docs/<feature-slug>/spec.md and pauses. It covers what you're building, why, and how, with explicit sections for design decisions, invariants, error behavior, and testing strategy. The spec is written for an AI agent to execute, so it emphasizes specificity over prose: testable requirements, concrete interfaces, version pinning for dependencies, and explicit choices where multiple implementations would diverge. Useful when you say "spec this out" or when a task has enough moving parts that you want to review the approach before committing to an implementation path.

skills.sh installs: 423
GitHub stars: 428
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add owainlewis/blueprint --skill spec --agent claude-code

Installs into .claude/skills of the current project.

Files

Spec

Fill in the spec template for one feature. The top half says what to build and for whom. The bottom half says how, for the agent that builds it. The spec is the ticket.

Process

Read the request, the code it touches, REQUIREMENTS.md, and ARCHITECTURE.md.
Ask the user about each choice that changes behavior, data, interfaces, or security. Give your recommended answer with each question.
Copy the template into the ticket, or into docs/<feature-slug>/spec.md if there is none. Fill every section, replace every bracketed placeholder, and delete the guidance comments. Write "None" where a section has nothing.
Compare the spec with REQUIREMENTS.md and ARCHITECTURE.md. Ask the user to settle each place they disagree.
Stop when another agent could build it without guessing a decision.