The Agent Skills Directory

[SAFE]: The skill implements a strictly controlled execution loop consisting of three phases: blueprinting, forging, and code review. It relies on standard internal tooling and state management for its operation.
[SAFE]: Data management is confined to local snapshots within a .rcs/ directory, following common development practices for maintaining task context.
[SAFE]: Execution policies include explicit stop conditions for credential-gated or destructive actions, ensuring human-in-the-loop oversight for high-risk operations.
[SAFE]: No external downloads, remote code execution patterns, or credential exfiltration risks were detected in the instructions.

autopilot