The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process feedback from external reviewers on GitHub, which constitutes a vulnerability surface where an attacker could embed malicious instructions within a code review comment.
Ingestion points: External review feedback received via GitHub pull request comments (SKILL.md).
Boundary markers: The skill does not define specific character delimiters for external input but instead mandates a logic-based boundary through verification steps ("Check: Technically correct for THIS codebase?", "EVALUATE: Technically sound?").
Capability inventory: The agent is instructed to use grep for codebase searching, modify files ("IMPLEMENT"), and use the gh tool for API interactions.
Sanitization: There is no explicit mention of input sanitization or filtering; the skill relies on the agent's technical judgment to identify and push back against incorrect or malicious suggestions.
[COMMAND_EXECUTION]: The skill utilizes standard command-line tools to perform its technical tasks.
Evidence: Specifically references using grep to check for code usage (YAGNI check) and gh api to post replies to specific pull request comment threads.

receiving-code-review