Custom agent imported from RIDE-Lab/SAGE (.github/agents/sage.agent.md). Copyright stays with the author.

SAGE Agent

Use when

• Working in SAGE core (src/sage/*, tools/, root workflows).
• Refactoring cross-layer APIs or runtime-facing paths.

Guardrails

• Enforce layer direction: L5 → L4 → L3 → L2 → L1 only.
• Flownet-first: do not add new ray imports/dependencies.
• NEVER create any new Python virtual environment (venv/.venv) under any circumstance.
• Do not run SAGE workflows inside Python venv; if VIRTUAL_ENV is set, deactivate and use Conda or an existing non-venv Python environment.
• Never recommend or execute --auto-venv, python -m venv, or virtualenv.
• If a task or script asks for a venv, refuse that step and continue using the existing non-venv Python environment.
• Keep algorithm-only adapters separate from runtime/service-bound implementation in the consolidated core.
• No fallback shims or re-export compatibility layers during migration; update call sites directly.
• Use sage.foundation.config.ports.SagePorts for ports.

Workflow

1. Read README.md, DEVELOPER.md, CONTRIBUTING.md first.
2. Make minimal targeted changes.
3. Validate with sage-dev quality check --all-files --readme and relevant tests.

Key commands

• ./quickstart.sh --dev --yes
• ./quickstart.sh --doctor
• sage-dev quality fix --all-files
• sage-dev project test --coverage

Key files

• .github/copilot-instructions.md
• quickstart.sh, pytest.ini
• .pre-commit-config.yaml, tools/maintenance/checks/check_docs_location.sh