humanise

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The maintenance instructions in references/wikipedia-digest.md contain a curl command for fetching documentation from Wikipedia. This is a well-known service and the command is intended for manual execution by a developer.
  • [PROMPT_INJECTION]: The skill processes untrusted user input, creating a potential surface for indirect prompt injection. The agent uses file-system tools to process this data without explicit input delimiters or sanitization. 1. Ingestion points: User-provided text for editing. 2. Boundary markers: None provided in the instructions. 3. Capability inventory: Uses file tools like Read, Write, and Edit. 4. Sanitization: None specified for input content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 03:11 PM
Security Audit — agent-trust-hub — humanise