Implemented three-layer defense mechanism to prevent AI-generated code from breaking the codebase:
- Layer 1: AGENTS.md with universal AI agent instructions
- Layer 2: Husky pre-commit hooks with test file checking
- Layer 3: GitHub Actions CI with TypeScript, ESLint, tests, and coverage
Includes design document, test infrastructure, and coverage thresholds.
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
- Update architecture diagram with SessionStore
- Add session management section to README
- Document IPC API for sessions
- Update system design with implemented SessionStore
- Resolve open questions (Q1, Q2, Q4, Q5)
- Add Q6 for future agent state restoration discussion
- Fix agent repository links
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Set up Electron + React + TypeScript + Tailwind CSS v4
- Implement Conductor for ACP agent communication
- Add AgentProcess for subprocess lifecycle management
- Create CLI test command (pnpm test:acp) with --cwd and --log options
- Support multiple agents: opencode, codex, gemini
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
