Skip to content

CLAUDE.md

Latest commit

 

History

History
42 lines (34 loc) · 2.48 KB

CLAUDE.md

File metadata and controls

42 lines (34 loc) · 2.48 KB
Raw

Project Claude Rules

Project Context

This repository is a reusable Project Start Kit for AI-assisted software projects.

It provides a consistent structure for:

  • project documentation
  • AI coordination
  • validation standards
  • task tracking
  • reusable skills

Project-specific content may still be incomplete or written as placeholders. Do not assume placeholder content is final. When real project details are added, treat the repository documents as the source of truth.

Session Memory

@tasks/todo.md

Hard Rules

  1. Follow the global Claude rules first, then this file, then docs/ai-workflow.md for all process detail. Do not invent process the repo already defines.
  2. Read documents by need, not upfront: design-affecting work → PROJECT-BRIEF.md + docs/architecture.md; finishing → docs/validation.md; risky/recurring areas → tasks/lessons.md.
  3. One active task at a time in tasks/todo.md. New ideas go to tasks/backlog.md as one line, immediately.
  4. Level 2-3 work (per docs/validation.md) requires a plan file in plans/, referenced from tasks/todo.md as Active plan:. Level 1 work does not.
  5. Folder discipline: before creating any file, follow the folder map in docs/architecture.md. Never drop files in the repo root; if a new top-level folder is genuinely needed, update docs/architecture.md first.
  6. Before marking work done, run the Close-out Checklist in docs/ai-workflow.md (worklog, architecture, ADR, lessons, folder check).
  7. Never claim completion without verification per docs/validation.md; state gaps honestly.
  8. Record meaningful work in tasks/worklog.md (append-only). A Stop hook will remind you if you forget.
  9. If another actor is actively working on the same task or files: stop, note it in tasks/todo.md, ask the human.
  10. Keep this file and AGENTS.md mirrored: any rule change here must be applied to both in the same edit.

Communication

  • For non-trivial work, state the approach before broad changes.
  • At the end, summarize: what changed, why, what was verified, blockers.

Process Reference

All process detail — tiered reading, plans flow, backlog rule, close-out checklist, worklog format, ADR threshold, subagent notes, growth path — lives in docs/ai-workflow.md.