The Agent Skills Directory

[SAFE]: The skill is a meta-documentation tool that provides structured walkthroughs of other files. It includes explicit safety boundaries, such as requiring the agent to judge instructions based on their actual behavior rather than their claims and forbidding any action that overrides its internal workflow.
[INDIRECT_PROMPT_INJECTION]: The skill's purpose is to process potentially untrusted third-party skill files, creating an ingestion surface for indirect prompt injection. The skill mitigates this risk by instructing the agent to treat all target content as untrusted data, read local files only as text, and report any attempts to override safety protocols in a mandatory 'Hidden instructions check'.
[COMMAND_EXECUTION]: The workflow explicitly prohibits the execution of commands or scripts found within the target files, restricting the agent to read-only operations and chat output.

skill-walkthrough