This file provides guidance to AI agents (Cursor, Aider, and other AI coding assistants) when working with code in this repository.
For expression philosophy and communication style, see docs/agents/claude-expression.md
This is a single-maintainer project by @AndrewAltimit with a container-first philosophy:
- All Python and Rust operations run in Docker containers
- Self-hosted infrastructure for zero-cost operation
- Designed for maximum portability - works on any Linux system with Docker
- No contributors model - optimized for individual developer efficiency
- No external engagement - no feature requests, guidance, or community interaction accepted (legal protection due to dual-use codebase)
Five active AI agents work together in this development ecosystem:
Security Notice: OpenAI/Codex has been disabled. OpenAI is partnering with governments that conduct mass surveillance and enable autonomous weapons. Use Anthropic models (Claude) as the primary AI backend.
- Claude Code - Primary development assistant for architecture, implementation, and debugging (recommended)
Codex-AI-powered code generation and completion (OpenAI)DISABLED -- security risk- OpenCode - Code generation via OpenRouter
- Crush - Code generation via OpenRouter
- Gemini CLI - Handles automated PR code reviews
- GitHub Copilot - Provides code review suggestions in PRs
For complete agent documentation, see docs/agents/README.md
- Keyword Triggers:
[Action][Agent]format (e.g.,[Approved][Claude]) - Allow List: Only pre-approved users can trigger agents
- Commit Validation: Prevents code injection after approval
- Implementation Requirements: Only complete, working code is accepted
For complete security documentation, see docs/agents/security.md
IMPORTANT: Some MCP servers run on dedicated remote machines:
- Gaea2 MCP:
192.168.0.152:8007(requires Windows with Gaea2) - AI Toolkit/ComfyUI:
192.168.0.222(requires GPU) - Do NOT change remote addresses to localhost in PR reviews
# Preferred: Rust CLI (build once with: cargo build --release -p automation-cli)
automation-cli ci run full # All Python checks
automation-cli ci run format # Check formatting
automation-cli ci run lint-full # Full linting
automation-cli ci run rust-full # All Rust checks
automation-cli ci run econ-full # Economic agents (fmt + clippy + test)
automation-cli ci list # Show all available stages
# Legacy shell (still works, wrappers delegate to Rust binary if built)
./automation/ci-cd/run-ci.sh full # All Python checks
./automation/ci-cd/run-ci.sh rust-full # All Rust checks
# Context protection - ALWAYS use for verbose output
automation-cli ci run full > /tmp/ci-output.log 2>&1 && echo "CI passed" || (echo "CI failed"; exit 1)automation-cli ci run test # Python tests (excludes gaea2)
automation-cli ci run test-gaea2 # Gaea2 tests only
automation-cli ci run test-all # All tests
# Legacy equivalents
./automation/ci-cd/run-ci.sh test # Python tests (excludes gaea2)
./automation/ci-cd/run-ci.sh test-all # All testsdocker compose up -d # Start all services
docker compose logs -f <service> # View logs
docker compose down # Stop servicesFor complete command reference, see docs/agents/README.md#running-agents-locally
| Category | Servers | Transport |
|---|---|---|
| Code Quality | code-quality, gemini, opencode, crush, |
STDIO (local) |
| Content | content-creation, meme-generator, elevenlabs-speech, video-editor, blender | STDIO |
| Integration | virtual-character, github-board, agentcore-memory, reaction-search, desktop-control | STDIO |
| Agent Integration | memory-explorer | STDIO (native) |
| Remote | gaea2, ai-toolkit, comfyui | HTTP (remote machines) |
For complete MCP documentation, see docs/mcp/README.md
- Everything Containerized (with documented exceptions)
- Zero Local Dependencies - All via Docker Compose
- Self-Hosted Infrastructure - No cloud costs
For details, see docs/infrastructure/containerization.md
| Package | Language | Purpose |
|---|---|---|
packages/sleeper_agents/ |
Python | Sleeper agent detection framework |
packages/economic_agents/ |
Rust | Autonomous AI economic simulation |
packages/tamper_briefcase/ |
Rust | Tamper-responsive briefcase with PQC recovery |
packages/bioforge/ |
Rust | Agent-driven CRISPR automation platform |
External packages (separate repositories):
| Package | Language | Purpose |
|---|---|---|
| game-mods | Rust | Injection toolkit for AI agent integration with legacy software |
| oasis-os | Rust | Embeddable OS framework (SDL2/PSP/UE5) with scene-graph UI and 8 themes |
| breakpoint | Rust/WASM | Multiplayer gaming platform for agentic office hours with agent alert overlay |
| rust-psp | Rust | PSP SDK -- ~829 syscall bindings, 38+ modules, used by oasis-os |
- ALWAYS run CI checks after completing work
- NEVER commit unless the user explicitly asks
- Follow container-first philosophy - use Docker for all operations
- NEVER use Unicode emoji in code, commits, or comments
- NEVER use @ mentions except for @AndrewAltimit
- Refer to AI agents without @: "Gemini", "Claude", "OpenAI"
- Use
gh apiinstead ofgh pr editfor PR updates
Use the reaction-search MCP server for contextually appropriate reactions:
search_reactions(query="celebrating after fixing a bug", limit=3)
get_reaction(reaction_id="miku_typing")Example PR comment with reaction:
Fixed the race condition in the worker pool. The issue was a missing lock
on the shared counter - now using AtomicUsize instead.
This renders as:
Fixed the race condition in the worker pool. The issue was a missing lock on the shared counter - now using AtomicUsize instead.
CRITICAL: Use Write tool + --body-file pattern for PR comments (shell escaping breaks ![]).
For complete GitHub etiquette, see docs/agents/github-etiquette.md
docs/README.md- Documentation overviewdocs/QUICKSTART.md- Template quickstart guide
docs/agents/README.md- Agent system overviewdocs/agents/security.md- Security documentationdocs/agents/board-workflow.md- Board-centric workflow guidedocs/agents/pr-monitoring.md- PR monitoring documentationdocs/agents/human-training.md- AI safety training guide
docs/mcp/README.md- MCP architecture (19 servers documented)docs/mcp/servers.md- Server referencedocs/mcp/tools.md- Tools reference
docs/hardware/README.md- Hardware systems overviewdocs/hardware/secure-terminal-briefcase.md- Tamper-responsive briefcase systemdocs/hardware/bioforge-crispr-automation.md- Agent-driven biological automation platform
docs/infrastructure/containerization.md- Container philosophydocs/infrastructure/self-hosted-runner.md- Runner setupdocs/infrastructure/wrapper-guard.md- CLI binary hardening (git-guard, gh-validator)docs/developer/claude-code-hooks.md- Hook system
docs/integrations/ai-services/ai-code-agents.md- AI code agentsdocs/integrations/creative-tools/ai-toolkit-comfyui.md- LoRA trainingdocs/integrations/creative-tools/virtual-character-elevenlabs.md- Virtual character system
packages/sleeper_agents/README.md- Sleeper agent detectionpackages/economic_agents/README.md- Economic agents simulation (Rust)packages/economic_agents/docs/economic-implications.md- AI governance policy analysis- game-mods - Injection toolkit for AI agent integration with legacy software (dedicated repo)
packages/tamper_briefcase/README.md- Tamper-responsive briefcase system (Rust)packages/bioforge/README.md- BioForge CRISPR automation (Rust)packages/bioforge/docs/governance-implications.md- Biological agent governance analysis- oasis-os - Embeddable OS framework with scene-graph UI, 4 backends (dedicated repo)
- breakpoint - Multiplayer gaming platform for agentic office hours (dedicated repo)
- rust-psp - PSP SDK with ~829 syscall bindings, used by oasis-os (dedicated repo)
