monorepo-workflows

# Monorepo Workflows Use this skill before broad repo changes, multi-service tasks, or when routing a task to the correct service. ## Primary Rules - Choose the service first, then the command surface. - Prefer root and workspace `package.json` scripts plus `bin/aicc.js`. - Treat `shell/` as convenience tooling with some path drift; verify before relying on it. - Do not edit generated output or local-only secret files. - Read nested `CLAUDE.md` files as you move into a service directory. ## Service Routing - API, auth, articles, comments, ratings, bias, sitewide chat backend, Pinecone sync: `backend/` - article browsing UI, auth UX, comments UI, ratings UI, SSE chat UI: `frontend/` - crawling, summarization, topic extraction, article ingestion jobs: `crawler/` - daily digest delivery and subscriber mail state: `newsletters/` - LangGraph pipeline and repo-local MCP server: `agentic_ai/` - Terraform, Kubernetes, monitoring, deploy wrappers: `infrastructure/` ## Workflow 1. Identify the owning service and read its local `CLAUDE.md`. 2. Confirm the current script or runtime entrypoint from `package.json`, `vercel.json`, Dockerfiles, or Makefiles. 3. Check whether the task has live side effects. 4. Make the smallest coherent change in source files, not generated output. 5. Run the narrowest useful validation for the touched area. ## References - For canonical commands and repo drift, read `references/commands-and-risks.md`. - For documentation entry points, read `references/docs-map.md`.