| name | trellis:guidelines |
| description | Creates or updates .specs/guidelines.md with project-wide stack, conventions, and principles. Use when starting a project or when a fundamental technology decision has changed. |
| allowed-tools | Read, Write, Edit, Glob, Grep, Bash(git:*), Task |
Guidelines
When to use
- "set up guidelines", "update guidelines"
- "change our stack", "add a convention"
- Any request to establish or modify project-wide principles
Create or update the project guidelines document at .specs/guidelines.md.
Recommended effort: medium. Interview-driven with structured generation into a fixed section template.
Pre-flight
Run python3 ${CLAUDE_PLUGIN_ROOT}/scripts/validate-prereqs.py guidelines and use the specsDir value from the JSON output. Abort if the output reports missing prerequisites.
Modes
This skill operates in two modes depending on whether .specs/guidelines.md already exists.
Create mode (no existing guidelines)
If .specs/guidelines.md does not exist, this is a fresh project. Ask the user:
- What is the tech stack? (language, runtime, framework, core libraries)
- What architectural style do they prefer? (functional, OOP, procedural, hybrid)
- Any hard conventions? (naming, file layout, error handling patterns)
- What's the testing philosophy? (when to test, what to skip, coverage attitude, framework, file location pattern, naming convention)
- Deployment and infrastructure constraints?
- What is the check command? This is the full CLI command chain that must pass for code to be considered correct — it runs after every implementation task. (e.g.,
npm run lint && npm run typecheck && npm run build && npm run test, or ruff check . && mypy . && pytest)
If the user provides a description or the project already has code, infer what you can and confirm with them.
Update mode (existing guidelines)
If .specs/guidelines.md already exists, read it first. Then ask the user:
- What changed? (new library, dropped dependency, revised convention, infrastructure shift)
- Why? (one sentence is fine; this helps evaluate downstream impact)
Generation
After gathering all user input via the interview, spawn the guideliner agent. Pass it: the specs directory path and all interview responses. The agent will generate the guidelines document.
Quality gate