analyzing-supply-chain-malware-artifacts

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The scripts/agent.py script performs metadata lookups on the official npm (registry.npmjs.org) and PyPI (pypi.org) registries. These are well-known services and the queries are used neutrally for integrity and provenance verification.
  • [PROMPT_INJECTION]: The skill handles untrusted data from package configuration files, presenting a surface for indirect prompt injection.
  • Ingestion points: Input files like package.json and setup.py are read and processed by scripts/agent.py.
  • Boundary markers: The script outputs its findings directly without using specific delimiters or safety instructions to separate analyzed data from the agent's context.
  • Capability inventory: The environment has access to standard filesystem operations and the Python requests library.
  • Sanitization: The skill performs static analysis using regular expressions and JSON parsing to flag known suspicious patterns (e.g., os.system, exec) rather than executing the untrusted content.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 03:56 PM