spec-creator

Create and maintain canonical design specifications as numbered markdown with JSON Schema sidecars.

Updated May 4, 2026
One-click install
npx skills add https://github.com/antstanley/skills --skill spec-creator-antstanley
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: spec-creator
Source: https://github.com/antstanley/skills/tree/main/plugins/spec-creator/skills/spec-creator
Command: npx skills add https://github.com/antstanley/skills --skill spec-creator-antstanley

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Codebases drift away from their documentation, and design decisions get lost. This Skill produces canonical design specifications that describe exactly what exists in the current branch, keeping architecture, domain models, and contracts documented in a consistent, cross-linked structure. ## Core Features & Use Cases - Canonical spec authoring: Investigates the actual code, then writes numbered, layered markdown spec pages (global and per-package) with a JSON Schema sidecar for typed entities. - Change spec lifecycle: Drafts single-document change specs under .specs/changes/ with a Proposed → Accepted → Implemented → Merged lifecycle, then merges them back into the canonical spec when code ships. - Companion skill orchestration: Delegates development-guidelines and design-guidelines pages to companion skills and enforces a mandatory cross-linking and checklist pass before handoff. - Use Case: Ask it to spec out a new app in a monorepo using an existing app's specs as a template; it reads the code, writes the numbered spec set, and updates the .specs/README.md index. ## Quick Start Ask the agent to create a spec for the app at apps/web using the existing specs in .specs/ as a template.

Frequently Asked Questions about spec-creator

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write a design spec for an existing codebase?

Point the skill at the package and optionally a template spec directory. It reads the code, plans a numbered file set (overview, domain model, component pages), writes each page describing what exists, and cross-links everything including the .specs/README.md index.

What is the difference between a canonical spec and a change spec?

A canonical spec is a numbered directory of pages describing what exists now, in present tense. A change spec is a single document under .specs/changes/ proposing a delta in future voice, with a Proposed → Accepted → Implemented → Merged lifecycle.

How do I merge a change spec after the code ships?

Tell the skill the change shipped. It applies each Proposed changes block to its canonical page, folds the JSON Schema fragment into canonical-types.schema.json, moves the file to .specs/changes/merged/ with Status Merged, and updates the index.

Does the spec support monorepos with multiple packages?

Yes. It uses a two-layer structure: global specs in .specs/ for cross-cutting concerns and per-package specs in .specs/<package>/specs/. Per-package pages open with a Read first pointer to global specs and document only deltas.

When should I not use this skill for documentation?

Skip it for READMEs, runbooks, tutorials, or API references. The skill writes structural design specs covering entities, lifecycles, contracts, and conventions, not prose-first how-to documentation.