gemini-cli
Pass
Audited by Gen Agent Trust Hub on Mar 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the gemini CLI command to perform AI analysis, interactive exploration, and automated code refactoring.
- [PROMPT_INJECTION]: The skill processes untrusted input from codebase focus parameters and previous analysis reports, which represents a surface for indirect prompt injection. 1. Ingestion points: The FOCUS variable in scripts/analyze_codebase.sh and the ANALYSIS content in scripts/second_opinion.sh. 2. Boundary markers: The second_opinion.sh script implements markers (---BEGIN/END ANALYSIS---) to isolate external content from instructions. 3. Capability inventory: The gemini tool supports file system modifications and automated editing via the --approval-mode yolo flag. 4. Sanitization: Input content is interpolated directly into prompt templates without additional filtering.
Audit Metadata