A lightweight, open-source safety framework to govern onboard AGI planners for space robotics (rovers, habitats, small satellites). It introduces a contract-based safety layer with (1) Safety Contracts defining pre/post conditions, budgets, and colli
Go to file
agent-7e3bbc424e07835b e130b8635e build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:44:10 +02:00
guardrail_space build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:44:10 +02:00
src/guardrail_space build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:44:10 +02:00
tests build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:44:10 +02:00
.gitignore build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:37:36 +02:00
AGENTS.md build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:44:10 +02:00
README.md build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:44:10 +02:00
guard_logs.jsonl build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:44:10 +02:00
pyproject.toml build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:44:10 +02:00
setup.py build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:37:36 +02:00
test.sh build(agent): new-agents-2#7e3bbc iteration 2026-04-19 22:44:10 +02:00

README.md

GuardRail.Space MVP

A lightweight, open-source safety framework to govern onboard AGI planners for space robotics.

  • Safety Contract DSL: define pre/post conditions, budgets, collision/warning rules, and trust policy.
  • Runtime policy engine: veto or adjust proposed actions.
  • Shadow planner: risk-aware alternative planning running in parallel.
  • Offline-first design: deterministic reconciliation and auditable logs.
  • Privacy-by-design: policy data stays on-device; aggregation only where allowed.
  • Lightweight verification hooks and adapters for CatOpt-style interoperability.
  • Gazebo/ROS integration path and HIL validation plan (skeletons provided).

This repository provides a minimal production-ready scaffold to be extended with adapters and a full test harness.