no-comments
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and modify external codebases and diffs provided by the user, which serves as an ingestion point for indirect prompt injection attacks. \n
- Ingestion points: The skill's Scope and Step 1 define the ingestion of the "caller's files or diff". \n
- Boundary markers: The instructions lack explicit requirements for using boundary markers or directives to ignore instructions embedded within the source code being reviewed. \n
- Capability inventory: The agent has the capability to modify the local file system (Steps 1, 4, and 5) and utilize sub-agent delegation via the
delegate_tasktool based on the results of the code analysis. \n - Sanitization: No sanitization or validation of the content within the user-provided files is described before the agent or its delegates process the data. \n- [COMMAND_EXECUTION]: The skill references and invokes several specialized platform tools to manage tasks and analyze code. \n
- Evidence: The instructions utilize
delegate_task,task_status, andtask_cancel(Step 1) for managing sub-agents, as well as operational commands such as/architect,/how, and/why(Steps 2, 3, and 5) to inform its implementation steps.
Audit Metadata