profiling-analysis-profiling-hostbound

Pass

Audited by Gen Agent Trust Hub on Apr 2, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns or security risks were identified in the skill content.
  • [PROMPT_INJECTION]: The skill exhibits a potential surface for indirect prompt injection as it processes external CSV data via the analyze_hostbound.py script. While no malicious intent is present, the following evidence chain is noted for risk assessment:
  • Ingestion points: The script scripts/analyze_hostbound.py accepts CSV files via the --input parameter.
  • Boundary markers: The provided documentation does not specify the use of delimiters or 'ignore embedded instructions' warnings for the data being analyzed.
  • Capability inventory: The skill is designed to perform data analysis and generate reports on system performance.
  • Sanitization: No information is provided regarding the validation or sanitization of the CSV content before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 2, 2026, 08:29 AM
Security Audit — agent-trust-hub — profiling-analysis-profiling-hostbound