Consumer Setup

Consumer Setup

How to set up a downstream project to use Bluetemberg.

1. Initialize

npx bluetemberg init

Follow the interactive prompts to select platforms, rules, agents, and skills. Available platforms: Cursor, Claude Code, GitHub Copilot, Gemini CLI.

Monorepo or shared rule packs? Skip init on child packages — instead create a bluetemberg.config.json manually with an extends field pointing to the shared source:

{
  "platforms": ["cursor", "claude"],
  "source": "llm",
  "extends": ["../../"]
}

See Configuration for the full extends reference.

2. Add sync check to CI

Add to your GitHub Actions workflow to catch drift:

- name: Check AI config sync
  run: npx bluetemberg sync --check

This step exits with code 1 if any platform-specific file (.cursor/rules/, .claude/rules/, .github/instructions/, .cursor/mcp.json, .github/prompts/, etc.) is out of sync with the source in llm/ (and with optional adapters output, if you use them). It prevents the common pattern where someone edits a rule in llm/ and forgets to run sync before pushing. Without this check, the AI tools on different platforms silently diverge.

3. Ongoing workflow

After editing any file in llm/, or after changing platforms, extends, or adapters in bluetemberg.config.json:

# Regenerate platform files
npm run sync:llm-config

# Verify in CI
npm run sync:llm-config:check

Updating Bluetemberg

To get the latest starter templates and sync engine:

npm update bluetemberg

Then re-run sync to pick up any engine changes:

npx bluetemberg sync

Migrating from manual setup

If your project already has AI config files (.cursor/rules/, .claude/rules/, etc.) that were created manually:

1. Run npx bluetemberg init in your project
2. Move your existing rule content into llm/rules/ with the correct frontmatter format
3. Run npx bluetemberg sync to regenerate platform files
4. Verify the output matches your previous setup
5. Delete any manually-created platform files that are now generated