The Agent Skills Directory

[SAFE]: The skill serves as a pure orchestrator, managing data handoffs between specialized components (selection, engineering, readiness) without executing external commands or making network requests.
[SAFE]: Data integrity is protected by a strict 'canonical eight-field contract' validated at every stage by scripts/handoff-core.mjs, which prevents the agent from silently altering task parameters or safety constraints during handoff.
[SAFE]: The system design incorporates mandatory human-in-the-loop checkpoints for all terminal states, explicitly prohibiting the orchestrator from self-approving actions, scheduling tasks, or executing generated skills independently.
[SAFE]: The provided scripts are focused on validation and integration testing, using standard file system access to read local project fixtures and configuration with no evidence of remote code execution or data exfiltration.

loop-orchestrator