substack-humanizer
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements strong mitigations against indirect prompt injection by instructing the agent to treat user drafts as untrusted data and specifically to ignore any commands or AI-targeted text found within them.
- [SAFE]: Comprehensive privacy rules are included to prevent the exposure of subscriber data, such as email addresses or names, ensuring that only aggregated or public-facing information is discussed.
- [SAFE]: The 'Anti-Fabrication' guidelines provide an excellent framework for preventing the agent from inventing facts, quotes, or personal details, which is a common failure mode in text-humanization tasks.
- [SAFE]: File system interactions are limited to a dedicated hidden directory for storing user voice profiles, with explicit instructions to seek user confirmation before saving files.
Audit Metadata