The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes PR comments which are untrusted external data. These comments are classified and then used to drive actions, including code fixes via the /do skill.
Ingestion points: PR comments are fetched from platforms like GitHub (SKILL.md).
Boundary markers: No explicit boundary markers or delimiters are defined to isolate untrusted comment text when passed to the classification or fix logic.
Capability inventory: The skill can invoke the /do tool (for code modification), the /loop tool (for persistence of the tending process), and PR platform actions such as merging or closing PRs.
Sanitization: There is no mention of sanitizing or escaping the content of PR comments before they influence the agent's decision-making process.
[COMMAND_EXECUTION]: The skill relies on executing other agent capabilities such as /do and /loop. While these are expected within the platform, they are triggered by the analysis of external, potentially adversarial PR comments.

tend-pr