case-analyzer
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill workflow ingests untrusted user data, such as chronological narratives and party descriptions, which are then interpolated into a structured legal report (CASE_ANALYSIS.md). This process lacks explicit boundary markers or sanitization logic, creating a surface where malicious instructions embedded within case facts could manipulate the agent's subsequent behavior or influence downstream tools like document drafters and archivers. Ingestion points: SKILL.md Step 1. Boundary markers: Absent. Capability inventory: Execution of local Python scripts and triggering of integrated research/drafting skills. Sanitization: Absent.
- [COMMAND_EXECUTION]: The skill instructs the agent to run local Python scripts (e.g., statute_of_limitations.py) using command-line arguments derived from user input, such as jurisdiction and incident dates. The provided invocation templates do not include sanitization or escaping guidance, posing a potential command injection risk if the agent's shell tool does not handle argument isolation securely.
Audit Metadata