gemini
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute the
geminicommand-line interface for code analysis tasks. It also provides maintenance instructions involving system utilities such asps(to monitor process state),lsof(to check network activity), andpkill/kill(to terminate hung processes). - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze large volumes of external data, including entire codebases and documentation sets (up to 1M tokens), which serves as a significant input surface for indirect prompt injection.
- Ingestion points: Codebase files and documentation directories specified via the
--include-directoriesflag inSKILL.mdandREADME.md. - Boundary markers: The instructions do not provide explicit delimiters or instructions for the model to distinguish between the primary task prompt and the untrusted data being analyzed.
- Capability inventory: The skill utilizes the
geminiCLI with a 'yolo' approval mode that enables automatic tool execution, alongside process management commands for the local shell. - Sanitization: There is no evidence of content sanitization, filtering, or escaping applied to the ingested codebase data before it is processed by the model.
Audit Metadata