Files
orchestration_engine/CLAUDE.md

22 lines
1.0 KiB
Markdown

# CLAUDE.md
## Direction
This repository holds two things. The **Execution engine** (`execution_engine/`) is the package an agent runs on. The **Orchestration engine** is the service every request comes through; it is not built yet.
Before designing or restructuring anything, read `CONTEXT.md` and `docs/adr/`. ADR-0001 is the anchor: the Execution engine names no Deliverable, no Cloud provider and no subject. Everything about a subject arrives in a Workflow.
The engine is built from a reference, `gcp_solution_architecture_agent`, which is only ever read (ADR-0002). Never edit a file there. Bring a module across with `scripts/bring.py`, which records where it came from in `REFERENCE.lock.json`.
Plans and history live in `.agents/plans/`, one directory per iteration; read `.agents/plans/README.md` before changing code. The design spec is `.scratch/engine/spec.md`.
## Checks
A step is done when all three pass:
```bash
.venv/bin/python -m pytest -q
.venv/bin/ruff check . && .venv/bin/ruff format --check .
.venv/bin/mypy .
```