anomaly-detector

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [Prompt Injection] (SAFE): No instructions attempting to bypass safety filters or override agent behavior.
  • [Data Exposure & Exfiltration] (SAFE): No hardcoded credentials, sensitive file access, or unauthorized network operations detected.
  • [Remote Code Execution] (SAFE): No remote code downloads or execution patterns found. Uses standard, trusted libraries including pandas, numpy, scikit-learn, and statsmodels.
  • [Obfuscation] (SAFE): No hidden or encoded content detected.
  • [Indirect Prompt Injection] (LOW): 1. Ingestion points: processes external data via pandas DataFrames/Series in scripts/anomaly_detection.py. 2. Boundary markers: absent. 3. Capability inventory: allowed tools include Bash, Read, Write, and Edit. 4. Sanitization: not performed on numerical data inputs. Analysis: standard data ingestion surface for analytical tools that could theoretically be used to pass instructions via data values if the agent interprets the results as commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:24 PM