humanizer
Warn
Audited by Gen Agent Trust Hub on Feb 15, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS] (MEDIUM): The installation process involves cloning a repository from an untrusted GitHub user ('blader'). Because this source is not within the trusted scope and the implementation details in SKILL.md are not available for review, the skill's safety cannot be verified.
- [PROMPT_INJECTION] (MEDIUM): The skill possesses an Indirect Prompt Injection surface (Category 8). It is intended to ingest untrusted external text. Ingestion points: Text provided via the '/humanizer' command or direct prompts. Boundary markers: None identified in the README. Capability inventory: As a Claude Code skill, it operates in an environment with file system and terminal access. Sanitization: No sanitization or filtering of input text is documented. Malicious instructions hidden in the text to be 'humanized' could potentially manipulate the agent's behavior.
Audit Metadata