adversarial-prompting

Pass

Audited by Gen Agent Trust Hub on Mar 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is instructed to automatically execute a local Python script scripts/export_analysis.py to save analysis results.
  • [PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection (Category 8).
  • Ingestion points: The skill processes user-provided problem descriptions and data to generate solutions and critiques.
  • Boundary markers: There are no specific delimiters or instructions defined to separate untrusted user input from the skill's operational logic.
  • Capability inventory: The skill possesses file-writing capabilities on the local system via the export script.
  • Sanitization: The export script performs basic alphanumeric sanitization for the output filename but does not sanitize or escape the content of the analysis before writing it to disk.
  • [DATA_EXFILTRATION]: The export script scripts/export_analysis.py targets a sensitive file path by writing files directly to the user's home directory (Path.home()). While intended for report generation, this interaction with a sensitive system path should be monitored.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 18, 2026, 09:55 AM