The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it is designed to parse and execute instructions found in the "Prompt for AI Agent" section of external GitHub PR comments.
Ingestion points: Comments are retrieved from GitHub PRs via the gh api command or the WebFetch tool.
Boundary markers: There are no explicit markers or system instructions to ignore potentially adversarial commands embedded within the fetched comment content.
Capability inventory: The skill is granted Bash, Edit, and Write capabilities, which could be misused if the agent obeys malicious instructions in a comment.
Sanitization: The skill does not perform sanitization or validation of the markdown/HTML content extracted from the comment body before the agent processes it.
[COMMAND_EXECUTION]: The skill utilizes the GitHub CLI (gh api) through the Bash tool to fetch pull request comments and metadata. This is a standard and expected use of system tools for this workflow.
[EXTERNAL_DOWNLOADS]: Content is fetched from GitHub PR URLs using WebFetch and API calls. These operations are directed at GitHub's infrastructure to retrieve the data necessary for the code review task.

sentry-code-review