The Agent Skills Directory

[SAFE]: The skill enforces a "Non-negotiable hardware boundary," which explicitly prohibits the agent from commanding physical equipment automatically and mandates a multi-step human operator safety gate for all live runs.
[SAFE]: Input validation logic in scripts/_common.py implements the safe_input_path function, which prevents path traversal attacks and symlink exploitation by resolving paths and verifying they remain within the current working directory.
[SAFE]: The skill utilizes strict schema enforcement and data sanitization for all processed JSON manifests and CSV transfer tables, including limits on file size, JSON nesting depth, and numeric bounds to prevent resource exhaustion or schema confusion.
[SAFE]: Bundled planning tools (e.g., check_deck_geometry.py, plan_transfers.py) are designed for offline, deterministic execution and do not perform network, serial, or USB operations.
[SAFE]: Dependency management is conducted through pinned installations of the legitimate PyLabRobot library (version 0.2.2) using the standard uv package manager.

pylabrobot