# AGENTS.md

Architecture and Collaboration Guide for CatOpt Studio MVP

- Scope: A minimal, production-ready scaffolding to bootstrap CatOpt Studio MVP and establish a repeatable workflow for future extensions.
- Tech stack (provisional): Python 3.11+, setuptools, pytest for tests. The MVP focuses on a clean DSL surface and packaging/tests scaffolding.
- Development workflow:
  - Create small, well-scoped changes that are testable and verifiable via `test.sh`.
  - Prefer small patches over large rewrites; ensure changes pass tests before progressing.
  - Use AGENTS.md as the contract for contribution and testing expectations.

- Testing and building:
  - `test.sh` should run: lint (if added), unit tests, and packaging build verification via `python3 -m build`.
  - Ensure packaging metadata and directory structure compile and can be installed locally.

- MVP Deliverables (high-level):
  1) Minimal DSL sketch (LocalProblem, SharedVariables, PlanDelta, PrivacyBudget, AuditLog).
  2) CatOpt Studio Graph wiring bootstrap (two starter adapters, e.g., rover_planner, habitat_module).
  3) Lightweight test harness and registry skeleton for conformance checks.
  4) Security and governance placeholders (DID or short-lived certs, per-message tags).
  5) Documentation and a ready-to-publish packaging description.

- Contributing rules:
  - Do not push to remote unless explicitly asked.
  - Follow the patch format and add tests for new functionality.
  - Update README and AGENTS.md if architecture changes occur.