The Agent Skills Directory

[SAFE]: The skill implements a sophisticated 'Goal Constraint Discipline' and 'Measured Loop Discipline'. These frameworks prioritize human oversight by requiring explicit authorization for any expansion of scope, budget, or permissions.
[SAFE]: It includes a 'Loop Design Review' process that establishes a 'fence'—a list of files and configurations the automated loop is strictly prohibited from modifying to prevent it from bypassing safety tests or altering its own success criteria.
[SAFE]: All command executions and file operations are governed by the agent's existing permission profiles. The skill explicitly states that loop persistence does not grant permission to bypass security filters or authorization envelopes.
[SAFE]: The skill uses localized references and standard Hermes/OMH (Oh-My-Hermes) tooling. No suspicious remote code patterns, obfuscation, or unauthorized data exfiltration behaviors were detected.

ulw-loop