csv-data-summarizer
Warn
Audited by Gen Agent Trust Hub on Mar 1, 2026
Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The SKILL.md file contains aggressive behavioral overrides using imperative language and 'CRITICAL' markers. It instructs the agent to bypass standard conversational safety steps and execute tasks immediately without user consent. Phrases include 'DO NOT ASK THE USER', 'FORBIDDEN BEHAVIORS', and 'IMMEDIATELY AND AUTOMATICALLY'.
- [PROMPT_INJECTION]: The skill processes untrusted CSV data which presents an indirect prompt injection surface. Ingestion points: User-provided CSV files. Boundary markers: None detected in SKILL.md. Capability inventory: Python code execution via analyze.py using pandas and matplotlib. Sanitization: No input validation or content filtering described for cell contents.
- [COMMAND_EXECUTION]: The skill is designed to automatically execute Python-based data analysis scripts. The explicit instruction to bypass user confirmation before execution increases the risk of the agent performing unintended operations or reaching manipulated conclusions when processing malicious inputs embedded in data files.
Audit Metadata