nature-geoscience

Pass

Audited by Gen Agent Trust Hub on Aug 27, 2026

Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXPOSURE]: The skill instructs the agent to access internal resource files using relative paths: ../../resources/source-basis.md and ../../resources/official-source-map.md. While these paths are restricted to the local environment, they represent an access point for internal configuration data.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external, untrusted content in the form of geoscience manuscripts.
  • Ingestion points: The agent evaluates manuscripts provided in the user prompt (SKILL.md).
  • Boundary markers: None explicitly defined to separate user content from system instructions.
  • Capability inventory: No scripts or dangerous capabilities (such as shell execution, network exfiltration, or file system writing) are included in the skill.
  • Sanitization: No specific sanitization or filtering logic is provided for the processed manuscript text.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 27, 2026, 02:06 PM
Security Audit — agent-trust-hub — nature-geoscience