The Agent Skills Directory

[SAFE]: No malicious behavior, obfuscation, or credential theft patterns were detected. The skill's operations are confined to stylistic text analysis and rewriting.
[SAFE]: The allowed tools (Read, Write, Edit, Grep, Glob, AskUserQuestion) are standard for file manipulation tasks and are used appropriately for reading and editing text provided by the user.
[SAFE]: The skill processes untrusted input (the text to be humanized), representing an indirect prompt injection surface. Analysis of the evidence chain: 1. Ingestion points: Text from conversation, arguments, or files (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Read, Write, Edit, Grep, Glob. 4. Sanitization: Absent. Despite the surface, the risk is assessed as safe because the instructions strictly bound the agent to linguistic rewriting and the skill does not use network tools for data exfiltration.

humanizer