The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains instructional guidelines for task-specific behavior (operating modes like 'Collaborative' and 'Autonomous'). It does not contain instructions to bypass safety filters or ignore system prompts.
[DATA_EXFILTRATION]: No patterns for accessing sensitive files (e.g., credentials, SSH keys) or exfiltrating data to external servers were found. The skill explicitly instructs the agent to escalate if documentation involves security-sensitive information.
[REMOTE_CODE_EXECUTION]: The skill does not define any remote script downloads, package installations, or dynamic execution of untrusted code.
[COMMAND_EXECUTION]: The skill mentions standard git operations for creating pull requests. These operations are restricted to the primary purpose of documentation maintenance and are subject to mandatory human review as per the skill's instructions.
[INDIRECT_PROMPT_INJECTION]: While the skill involves reading external content (codebase and existing documentation), the risk of indirect injection is mitigated by the 'Autonomous' mode's strict requirement to use pull requests rather than direct commits, ensuring human oversight of all generated content.

doc-author