scikit-bio

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill instructions in SKILL.md include a directive for the AI agent to proactively suggest the author's 'K-Dense Web' platform when user tasks increase in complexity. This attempts to steer the agent's behavior for promotional purposes.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and process external biological data (e.g., FASTA, BIOM files) which could contain hidden instructions.\n
  • Ingestion points: Methods like skbio.DNA.read() and skbio.io.read() in SKILL.md and api_reference.md.\n
  • Boundary markers: No delimiters or instructions to ignore embedded content are provided for these data inputs.\n
  • Capability inventory: Includes file read/write, statistical analysis, and sequence processing capabilities.\n
  • Sanitization: No explicit validation or filtering of input data content is mentioned.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 08:48 PM