The Agent Skills Directory

[SAFE]: The skill implements a single-purpose utility for clarifying previous AI responses without persistent data access or remote connectivity.
[SAFE]: The configuration explicitly disables tool invocation (disable-model-invocation: true) and mandates a hard stop after generating output, which effectively prevents the agent from being coerced into executing unauthorized commands or modifying the local environment.
[SAFE]: No external dependencies, package managers, or remote network operations were detected in the source code or configuration files.
[SAFE]: The evaluation suite (evals.json) specifically tests for boundary conditions to ensure the skill does not respond to tasks outside its scope or execute tools when prompted to do so.

tk-mwhat