Container wrapper that runs AI coding agents (Claude Code, Cursor, Gemini CLI) in isolated, secure containers.
make test # Unit tests (pytest)
make test-all # Unit + integration tests
make lint # Ruff linter
make typecheck # Mypy strict mode
make format # Auto-format with ruff
make build # Build container images locally
make clean && make run # Rebuild after container changes- All new features must include tests. Add tests in
tests/test_<module>.py; CLI flag tests go intests/test_cli.py. - After any code change, run
make test,make lint, andmake typecheck. Do not commit if any fail. - When adding or changing user-facing features, update
README.mdandsrc/paude/cli/help.py. - Use agent-agnostic language in user-facing text. Say "the agent" not "Claude" (except in agent-specific examples).
- When adding scripts to
containers/paude/, also add them to[tool.hatch.build.targets.wheel.force-include]inpyproject.tomlso they get bundled into the installed package. - When discovering unrelated bugs or tech debt, add them to
KNOWN_ISSUES.mdand continue with the current task.
src/paude/-- Python package. Key areas:agents/(agent definitions),backends/(podman and openshift),cli/(typer CLI),config/(configuration),container/(container management),transport/(local/SSH execution).containers/paude/-- Main container Dockerfile and entrypoints.containers/proxy/-- Network filtering proxy container.tests/-- Pytest test suite.
- Coding Standards -- file/method size limits, refactoring triggers, shared utility locations
- Security Model -- mount restrictions, credential handling, attack vectors
- Configuration -- config file format and resolution
- Remote Sessions -- SSH-based remote execution
- OpenShift Backend -- Kubernetes deployment
- Orchestration -- multi-agent workflows
- Sessions -- session lifecycle