research-platform-engineering

Pass

Audited by Gen Agent Trust Hub on Jun 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill scripts and instructions are well-defined and perform only the intended statistical analysis. There is no evidence of obfuscation, remote code execution, or credential theft.
  • [PROMPT_INJECTION]: The skill processes external data from a CSV file, which represents a surface for indirect prompt injection.
  • Ingestion points: scripts/research_platform_engineering_diagnostics.py reads untrusted data from input.csv.
  • Boundary markers: No explicit instruction delimiters are provided.
  • Capability inventory: The agent executes a Python script to calculate statistics and subsequently reads the resulting JSON report.
  • Sanitization: Data is sanitized by converting columns to numeric and datetime types using pandas, which prevents the execution of embedded text instructions within the data fields.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 17, 2026, 01:29 PM
Security Audit — agent-trust-hub — research-platform-engineering