idea19-missionledger-verifi.../AGENTS.md

1.7 KiB

MissionLedger Agents Guide

Overview

  • MissionLedger is a production-grade core for verifiable, contract-driven planning in intermittently-connected fleets.
  • This repo contains a minimal but production-oriented MVP to bootstrap the ecosystem: a DSL for per-asset LocalProblems, safety contracts, a shadow planner, a lightweight safety verifier, and a governanced ledger.

Architecture (high level)

  • missionledger.dsl: Defines the LocalProblem, SafetyContracts, PrivacyContracts, and PlanDelta data models.
  • missionledger.planner: ShadowPlanner that proposes a safe alternative plan in parallel to the executor.
  • missionledger.safety: Lightweight safety verifier used by both the main planner and shadow planner.
  • missionledger.governance: Tamper-evident governance ledger with hash chaining for auditability.
  • missionledger.adapter (placeholder): SDK adapters for various assets and runtime stacks (rover, drone, habitat, orbiters).

Tech stack (current MVP)

  • Python 3.8+ with a small, typed core using dataclasses.
  • No external runtime dependencies for MVP; easy to extend with real solvers later.

Development commands

  • Run tests: bash test.sh
  • Build the package: python3 -m build
  • Lint/type checks: not included in MVP yet; plan to add flake8/ruff and mypy in future iterations.

Tests and verification

  • The test suite exercises the LocalProblem DSL, the ShadowPlanner, the safety verifier, and the governance ledger entries.
  • Extend tests to cover full end-to-end delta-sync, proofs, and HIL adapters in later phases.

Contribution guidelines

  • Follow the architecture layout described above.
  • Add tests for any new feature and ensure tests pass locally before proposing PRs.