humanizer-ng

Pass

Audited by Gen Agent Trust Hub on Jul 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface. It is designed to ingest and process untrusted user-provided writing samples while possessing capabilities to read, write, and edit files.
  • Ingestion points: Input text for humanization and editing (SKILL.md).
  • Boundary markers: Not present; the instructions do not provide delimiters to separate user data from agent instructions.
  • Capability inventory: Read, Write, Edit, and AskUserQuestion tools.
  • Sanitization: No sanitization or validation of the input content is defined in the skill instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 17, 2026, 03:33 PM
Security Audit — agent-trust-hub — humanizer-ng