nature-cell-biology

Pass

Audited by Gen Agent Trust Hub on Sep 24, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: No malicious patterns detected. The content is exclusively focused on academic journal submission criteria and framing.
  • [NO_CODE]: The skill consists entirely of natural language instructions and does not include scripts, command execution, or external dependencies. It references local Markdown files (../../resources/source-basis.md) as a way to access ground-truth submission data, which is a benign configuration practice.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided manuscript details.
  • Ingestion points: User-provided manuscript abstracts, titles, and descriptions are processed via natural language interaction.
  • Boundary markers: None explicitly defined in the prompt instructions to separate user data from instructions.
  • Capability inventory: The skill possesses no subprocess calls, file-write, or network operations.
  • Sanitization: None present; however, the lack of system-impacting capabilities renders the injection surface safe.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 24, 2026, 11:33 AM
Security Audit — agent-trust-hub — nature-cell-biology