Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Spec Compiler Skill

Use this skill through an AI agent prompt to initialize, create, update, validate, compile, search, analyze, and maintain an AI-friendly Markdown specification corpus.

Users normally do not run the bundled scripts directly. Ask the agent to use $spec-compiler; the agent will inspect the project, run deterministic helpers when useful, edit Markdown files, validate results, and summarize what changed.

Main Prompt Examples

Initialize a project:

Use $spec-compiler to initialize this repository as a spec project.
Inspect existing docs first, create the local configuration and backlog,
and tell me what was created.

Create a canonical spec:

Use $spec-compiler to create a new API spec for "Refresh access token".
Ask only for missing required information, create the canonical Markdown doc
under specs/, validate it, and compile generated outputs.

Update existing specs:

Use $spec-compiler to update the auth component spec with the new dependency
on the token refresh API. Preserve existing IDs and links, then validate and
summarize affected files.

Analyze impact:

Use $spec-compiler to analyze what is affected if the auth API changes.
Return impacted document IDs, relation paths, and any active backlog items
that should be reviewed.

Pack implementation context:

Use $spec-compiler to pack context for implementing token revocation.
Include only the smallest useful spec sections, related facts, active backlog
items, and omitted context that may matter.

Default Locations

  • specs/: canonical target specification documents.
  • specs/shared/: shared or common-detail specifications.
  • .speccompiler/generated/: generated JSONL, backlinks, indexes, and relation data.
  • BACKLOG.md: project-local backlog.
  • spec-assets/: project-local templates, type definitions, retrieval recipes, and feedback rules.
  • speccompiler.json: project-local configuration.

Generated files can guide edits, but source-of-truth changes belong in specs/, specs/shared/, or project-local assets.

Skill Files

  • SKILL.md: instructions loaded by the AI agent.
  • scripts/speccompiler.py: deterministic helper used by the agent.
  • assets/templates/: starter Markdown templates.
  • assets/type-definitions/: baseline document type metadata.
  • references/: detailed product requirements loaded only when needed.
  • LICENSE: GNU Affero General Public License v3 text.

About

spec markdown editing helper skill

Resources

Stars

5 stars

Watchers

2 watching

Forks

Releases

Packages

Contributors

Languages