The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch and process pull request review comments using the gh CLI. Since these comments are external data provided by third parties, they could contain malicious instructions designed to manipulate the agent's behavior.
Ingestion points: The GraphQL query in SKILL.md used by the gh api command fetches the body content of review threads.
Boundary markers: The skill does not provide delimiters or instructions to treat the fetched comment text strictly as data.
Capability inventory: The agent is granted access to gh for network-based API interactions (including write operations like replying to and resolving threads) and git for local repository operations.
Sanitization: No logic is provided to sanitize, validate, or filter the text of the review comments before the agent processes them.
[COMMAND_EXECUTION]: The skill utilizes the GitHub CLI (gh) and git commands to perform legitimate development tasks, such as querying pull request metadata, enumerating workflow runs, and verifying the merge status of branches.

pull-request