The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the git and gh (GitHub CLI) tools to perform repository operations, manage branches, and interact with the GitHub API. These operations include fetching branches, committing changes, and merging pull requests, which are standard for the tool's intended purpose. Use of the --admin flag for merging is correctly restricted to the user's own pull requests.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file utilizes dynamic context injection (e.g., !gh repo view, !git branch --show-current) to gather environmental information when the skill is loaded. These commands are limited to read-only repository and user metadata and are used to inform the agent's context rather than execute untrusted input.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from GitHub PR reviews and comments.
Ingestion points: Data enters via gh pr view and GraphQL queries for reviewThreads nodes (referenced in monitor.md).
Boundary markers: None explicitly defined in the prompt instructions, though logic is compartmentalized by delegating triage to specialized auxiliary skills.
Capability inventory: The skill can perform file-system writes (via git), network operations (via gh), and state changes such as merging pull requests.
Sanitization: The skill employs jq -Rs to safely encode comment bodies into JSON before sending them to the GitHub API, preventing potential command injection or formatting issues from untrusted strings.

ship-pr