adaptyv

Pass

Audited by Gen Agent Trust Hub on May 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill communicates with the Adaptyv platform API hosted on AWS Lambda (lambda-url.eu-central-1.on.aws) and a protein solubility prediction service from the Technical University of Denmark (services.healthtech.dtu.dk). These are legitimate endpoints necessary for the skill's primary function.
  • [COMMAND_EXECUTION]: Instructions are provided for installing standard scientific and bioinformatics Python packages using the uv manager. These include biopython, pandas, torch, fair-esm, and specialized tools like soluprot and soluble-mpnn. All packages are established tools in the bioinformatics domain.
  • [PROMPT_INJECTION]: The skill processes external data, creating an ingestion surface.
  • Ingestion points: Protein sequences in FASTA format and user-provided webhook URLs for experiment completion notifications in SKILL.md and reference/examples.md.
  • Boundary markers: The reference/examples.md file includes a validate_fasta function designed to ensure input sequences only contain valid amino acid characters.
  • Capability inventory: Uses the requests library to communicate with remote APIs and the open function to save result files locally.
  • Sanitization: Implements character-level validation for protein sequences to prevent malformed data processing.
  • [SAFE]: No evidence of credential exfiltration, persistence mechanisms, or obfuscated content was found. The use of environment variables for API key management follows standard security best practices.
Audit Metadata
Risk Level
SAFE
Analyzed
May 28, 2026, 12:51 AM