What problem does it solve? Autonomous loop work can violate safety boundaries, skip human approval gates, or act while the system is paused. This Skill loads and enforces the AutoResearch loop constraints before any triage, editing, remote execution, or external writes so every action respects role scope and denylists. ## Core Features & Use Cases - Constraint Loading: Reads loop-constraints.md, docs/safety.md, and LOOP.md before any loop work begins. - Pause Detection: Exits without action when a .loop-paused file exists or STATE.md reports a paused status. - Gate and Denylist Enforcement: Counts and reports active human gates and denylist categories, applying them to every subsequent action. - Escalation Handling: Returns ESCALATE_HUMAN with the blocked action and required approval when authority is missing. - Use Case: Before an agent edits code, pushes to GitHub, or runs commands on a remote NPU server, invoke this Skill to verify the action is within role scope and not on the denylist. ## Quick Start Ask the agent to load and enforce the loop constraints before performing any triage, code edits, remote execution, or Git and connector writes.
