humanizer
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFENO_CODE
Full Analysis
- NO_CODE (SAFE): The skill is comprised entirely of Markdown (.md) files providing documentation and transformation rules. No Python scripts, Node.js code, or shell commands are included in the skill package.\n- Signage Analysis (SAFE): The content is based on legitimate Wikipedia editorial guidelines (WikiProject AI Cleanup) for identifying machine-generated text. The patterns provided are strictly stylistic and linguistic in nature, intended for text refinement.\n- Indirect Prompt Injection (SAFE): The skill possesses the surface for indirect prompt injection as it processes untrusted user input and allows tools with file system write permissions (Write, Edit). However, the instructions are explicitly limited to linguistic pattern replacement and lack any dangerous execution logic.\n
- Ingestion points: Input text or local files provided by the user for the humanization process.\n
- Boundary markers: None present in the skill instructions to delimit untrusted data.\n
- Capability inventory: Read, Write, Edit, Grep, Glob, and AskUserQuestion tools allow for file discovery and modification.\n
- Sanitization: None present; the skill relies on the underlying agent's safety layers for handling untrusted data.
Audit Metadata