The Agent Skills Directory

External Content Processing: The skill ingests pull request comments and workflow logs from GitHub via the gh CLI (see scripts/gh_pr_watch.py). While this is necessary for monitoring PR feedback, it means the agent processes text provided by other users. The skill includes logic to prioritize comments from repository owners, members, and collaborators to help manage this interaction safely.
Command Execution via Subprocess: The Python scripts utilize the GitHub CLI (gh) to fetch PR metadata and trigger workflow reruns. These system calls are implemented using structured argument lists, which is a standard and secure practice that prevents shell-based injection vulnerabilities.
State Management: The skill tracks monitoring progress using JSON state files stored in the /tmp directory. This mechanism ensures that the agent handles each piece of feedback only once, preventing redundant actions or comments.

babysit-pr