The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources including human reviews, bot reviews, PR comments, and CI failure logs to drive its automated fixing and responding logic.
Ingestion points: Human reviews, bot reviews, human comments, and CI failure output logs are read and acted upon as described in SKILL.md.
Boundary markers: The skill relies on external logic defined in a dependent 'git' skill (specifically review-responses.md) to handle responses, but the provided instructions lack explicit delimiters or 'ignore' instructions to isolate untrusted data from the agent's control flow.
Capability inventory: The agent has the capability to execute shell commands (gh CLI), run Python scripts, and is specifically instructed to "fix failures" and "push" changes, providing an execution path for injected instructions.
Sanitization: No explicit sanitization or filtering of external comment content is specified in the instruction set.
[COMMAND_EXECUTION]: The skill uses the GitHub CLI and other tools to manage the PR lifecycle.
Evidence: Execution of gh pr view --json number,url,headRefName,state,isDraft to resolve target PR details.
Evidence: Reference to running a flick command to manage PR draft status.
Evidence: Execution of python3 <git-skill-dir>/scripts/pr-watch.py with various arguments for the monitoring loop.
[DYNAMIC_EXECUTION]: The skill executes a Python script located at a path determined at runtime from a dependent skill.
Evidence: The instruction python3 <git-skill-dir>/scripts/pr-watch.py uses a placeholder <git-skill-dir> which is resolved dynamically when the environment loads the required 'git' skill.

watch-pr