The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from GitHub PR comments and reviews, which are untrusted external sources.
Ingestion points: The skill fetches PR state including comments and reviews using the gh pr view command in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the text of these comments as data rather than instructions, nor are there warnings to ignore embedded instructions.
Capability inventory: The agent possesses significant capabilities, including file modification, committing code, force-pushing to branches, and merging PRs via the gh CLI as described in the triage and loop steps of SKILL.md.
Sanitization: The skill lacks sanitization or validation of the ingested comment prose before the agent evaluates it for potential action.
[COMMAND_EXECUTION]: The skill relies extensively on the gh (GitHub) CLI to perform actions.
The agent executes gh pr view, gh run view, gh pr checks, and gh pr ready to manage the PR lifecycle.
While these are standard tools, the integration with the loop skill for automated pacing increases the impact of any potentially misdirected commands.

babysit