humanizer

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting arbitrary text from a user or external source and processing it using a set of instructions to rewrite AI patterns. This exposes an attack surface for indirect prompt injection where instructions embedded in the input data could be executed by the agent.
  • Ingestion points: User-provided text submitted for editing via the /humanizer command or direct prompt, as documented in SKILL.md.
  • Boundary markers: Absent. The instructions do not define delimiters (such as specific XML tags or triple backticks) or include explicit directions for the agent to ignore instructions found within the input text.
  • Capability inventory: The skill is configured in SKILL.md to use the Read, Write, Edit, Grep, and Glob tools, enabling the agent to modify files in the environment based on its analysis of the input.
  • Sanitization: Absent. The skill does not perform any validation, filtering, or escaping of the input text before it is processed by the model.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 03:58 AM