benchling-integration

Pass

Audited by Gen Agent Trust Hub on Oct 24, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill utilizes environment variables for all sensitive credentials, such as API keys and OAuth secrets, adhering to security best practices for secret management. Access is controlled via environment variables like BENCHLING_API_KEY and BENCHLING_CLIENT_SECRET.
  • [SAFE]: Network operations are confined to the official Benchling API (benchling.com) and well-known academic domains (arxiv.org). The use of the benchling-sdk is consistent with the skill's stated purpose of facilitating lab data automation.
  • [INDIRECT_PROMPT_INJECTION]: The skill contains logic for processing external scientific data (FASTA files) and Benchling API responses, creating a potential surface for indirect prompt injection if those sources are attacker-controlled.
  • Ingestion points: FASTA file parsing in SKILL.md and sequence listing in sdk_reference.md.
  • Boundary markers: Not explicitly defined in the provided prompts to isolate data from instructions.
  • Capability inventory: File system writes via Python open and Benchling API mutations across all SDK scripts.
  • Sanitization: No explicit sanitization logic is shown for sequence metadata or field values before they are processed or stored.
  • [SAFE]: All dependencies, including benchling-sdk and biopython, are specified with pinned versions and are standard, reputable libraries in the bioinformatics domain.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 24, 2026, 09:46 PM