humanize
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill processes user-provided text and writing samples, which inherently creates an attack surface for indirect prompt injection. While the instructions are focused on stylistic analysis, malicious commands embedded within the processed text could theoretically influence the agent's behavior.
- Ingestion points: User-provided text for humanization and optional file paths for style samples (SKILL.md).
- Boundary markers: The skill does not implement specific delimiters or 'ignore' instructions to wrap external content during interpolation.
- Capability inventory: The agent has access to file editing and system inspection tools (Read, Write, Edit, Grep, Glob).
- Sanitization: No sanitization or safety-filtering is applied to the input text prior to processing.
Audit Metadata