protein-sequence-msa

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill communicates with the European Bioinformatics Institute (EBI) Clustal Omega REST API at www.ebi.ac.uk to perform protein sequence alignments. This is a well-known and trusted scientific service, and the interaction is handled via a dedicated client that respects the service's terms of use.
  • [COMMAND_EXECUTION]: The skill uses a local Python wrapper script (scripts/msa_align.py) to manage API requests and file operations. This script includes basic validation for file size (4MB limit) and sequence count (4000 limit) before execution.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted sequence data from local FASTA files and retrieves alignment results from an external API, which are then analyzed by the agent.
  • Ingestion points: scripts/msa_align.py reads user-provided FASTA files and the text-based alignment results returned by the EBI API.
  • Boundary markers: No specific delimiters or "ignore instructions" warnings are utilized when processing the sequence content or alignment output.
  • Capability inventory: The skill possesses network access (to the EBI API), local file read capabilities (for input sequences), and local file write capabilities (for storing alignment results).
  • Sanitization: While the script validates the format and size of the input, it does not sanitize sequence headers or content for potential instruction-like strings; however, the risk is minimized by the scientific context of the analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:46 AM