data-exploration-visualization

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • SAFE (SAFE): No malicious patterns such as obfuscation, exfiltration, or persistence mechanisms were detected. The skill uses well-known data science libraries (pandas, scikit-learn, etc.) and follows legitimate patterns for exploratory data analysis and report generation.
  • Indirect Prompt Injection (LOW): The skill is designed to ingest and analyze external data files, which creates an attack surface for indirect prompt injection. 1. Ingestion points: The skill loads external data via EDAAnalyzer.load_data as specified in README.md and SKILL.md. 2. Boundary markers: No boundary markers or 'ignore embedded instructions' warnings are present in the documentation. 3. Capability inventory: The skill is granted access to high-privilege tools including Bash, Write, Read, Glob, and Grep as defined in SKILL.md. 4. Sanitization: There is no evidence of sanitization or validation of the content within processed data files before it influences agent actions.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 05:56 PM