unibind-database

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches transcription factor binding site information and metadata from the UniBind database hosted at unibind.uio.no.
  • [COMMAND_EXECUTION]: Executes the unibind_api.py wrapper script using the uv package manager to facilitate database queries and file downloads.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from an external scientific database, which represents an attack surface for instructions embedded in processed data.
  • Ingestion points: JSON API responses and downloaded genomic coordinate/sequence files (BED/FASTA).
  • Boundary markers: The skill instructions recommend using tools like jq or jp to filter and extract specific fields from large datasets, which limits the volume of external data entering the agent's context.
  • Capability inventory: The skill performs network requests for API communication and writes retrieved data to the workspace file system.
  • Sanitization: Standard JSON parsing and binary writing are used; no specific content validation or sanitization is performed on the biological data fields.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:46 AM
Security Audit — agent-trust-hub — unibind-database