The Agent Skills Directory

[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it processes untrusted code changes and passes them to an LLM-based tool for review.\n * Ingestion points: The changed_files list in SKILL.md is passed to the review tool.\n * Boundary markers: There are no delimiters or "ignore" instructions in the message sent to repoprompt/chat_send.\n * Capability inventory: The skill executes the repoprompt/chat_send MCP tool.\n * Sanitization: No validation or sanitization is performed on the content of the changed files.\n- [SAFE]: The skill utilizes the repoprompt MCP tools (manage_workspaces, chat_send) to perform code reviews.\n- [SAFE]: The skill uses a "fail-open" approach in its error handling logic, where failures in the toolset lead to an automatic approval verdict to prevent blocking development workflows.

rp-reviewer