alterlab-hmdb

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses a Python script (scripts/query_hmdb.py) to fetch metabolic data from the Human Metabolome Database at https://www.hmdb.ca. This is a well-known academic resource, and the operations are consistent with the skill's stated purpose. Documented neutrally as it targets a reputable domain.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external XML records, which presents a potential surface for indirect prompt injection.
  • Ingestion points: scripts/query_hmdb.py fetches XML data from hmdb.ca and parses it into a JSON structure.
  • Boundary markers: The script produces structured JSON output using json.dumps, providing clear structural separation between the retrieved data and the agent's instructions.
  • Capability inventory: The skill utilizes standard library networking and XML parsing in scripts/query_hmdb.py, with execution allowed via the Bash(python:*) tool defined in the metadata.
  • Sanitization: The script uses robust standard library XML parsing (xml.etree.ElementTree) and structural JSON encoding to handle external content securely.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 02:36 PM