The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the gh (GitHub CLI) and git command-line tools to interact with repositories.
Evidence: SKILL.md describes using gh pr view, gh pr review, and git push.
Evidence: scripts/fetch_comments.py uses subprocess.run to execute gh api graphql for detailed comment retrieval. These are standard operations for a developer tool and use the official CLI interface.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data (GitHub comments) which could theoretically contain instructions intended to influence the agent's behavior.
Ingestion points: scripts/fetch_comments.py (reads GitHub GraphQL API response).
Boundary markers: The skill instructs the agent to '编号' (number) and '拟修复逻辑' (propose fix logic) for user confirmation before acting, which provides a human-in-the-loop safety boundary.
Capability inventory: The agent can modify local files, commit code, and push to remote branches.
Sanitization: No specific technical sanitization is mentioned, but the workflow requires interactive task confirmation (Clarification step in SKILL.md), mitigating the risk of automatic execution of malicious instructions found in comments.

gh-address-comments