The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data (GitHub issue comments, PR bodies, and git logs) to generate summaries and responses, which creates a surface for indirect prompt injection.
Ingestion points: gh issue view, gh pr view, git diff, and git log commands across multiple reference files including references/comment-issue.md, references/update-pr.md, and references/commons.md.
Boundary markers: The skill employs single-quoted HEREDOC syntax ('EOF') to prevent shell expansion of generated content and requires explicit user confirmation via AskUserQuestion before performing any write operations.
Capability inventory: The skill uses gh and git CLI tools to create, update, and comment on GitHub issues, pull requests, and discussions.
Sanitization: The skill relies on the agent's internal filtering and user oversight for the content of generated text; it implements robust shell-level sanitization via HEREDOC quoting to prevent command injection from the generated content.

yeet