The Agent Skills Directory

[DATA_EXFILTRATION]: The skill utilizes the official gh CLI to interact with GitHub, a well-known service, for fetching PR data and posting comments. These actions are aligned with the skill's primary purpose and do not constitute unauthorized exfiltration.
[PROMPT_INJECTION]: The skill processes external data from PR diffs and comments, creating a potential surface for indirect prompt injection. This is mitigated through structured sub-agent prompts and clear boundary instructions.
Ingestion points: Data retrieved via gh pr diff, gh pr view, and local source file reads.
Boundary markers: The sub-agent is explicitly instructed to 'Review ONLY the changes in this PR diff' in github-pr-workflow.md.
Capability inventory: Includes reading local files and writing to GitHub via gh pr comment in github-pr-workflow.md.
Sanitization: The skill uses shell heredocs in github-pr-workflow.md to wrap external strings, preventing command injection when calling the GitHub CLI.
[COMMAND_EXECUTION]: Employs standard git and gh commands to perform its core functions. These operations are expected for a development tool and follow safe implementation practices.

review