tooluniverse-antibody-engineering

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to the processing of untrusted user input combined with powerful capabilities.\n
  • Ingestion points: Antibody VH/VL sequences provided as text or FASTA format, and target antigen names, entering the context in Phase 1 through Phase 6.\n
  • Boundary markers: The prompt instructions do not include explicit delimiters or 'ignore instructions' warnings for the sequence data ingested from the user.\n
  • Capability inventory: The skill is capable of performing network requests through scientific tools (IMGT, AlphaFold, UniProt, STRING) and writing optimization reports and sequence files to the local file system.\n
  • Sanitization: No sanitization, validation, or instruction-filtering is applied to the antibody sequences before they are used in tool operations or documentation generation.\n- [COMMAND_EXECUTION]: The skill involves the generation and execution of local Python scripts to implement the engineering pipeline.\n
  • Evidence: The workflow details provide code templates that the agent is expected to implement and run (e.g., antibody_pipeline.py).\n
  • Analysis: This script generation follows known templates for data processing and is a core part of the skill's intended functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 11:40 PM