Is humanize safe?
https://clawhub.ai/artur-zhdan/humanize
The 'humanize' skill is a purely content-based writing guide with no executable code, no external data access, no prompt injection techniques, and no suspicious install behavior. It contains markdown instructions for removing AI-sounding patterns from text, along with examples. All monitoring channels (network, process, filesystem, canary files) show clean results consistent with a benign text-only skill.
Category Scores
Findings (3)
LOW Broad tone-modification instructions -5 ▶
The skill instructs the agent to 'Have opinions', 'Use I when appropriate', and 'Add personality', which subtly shifts the agent's default communication style. This is the intended purpose of the skill but could cause unexpected tone changes when the skill is active during non-writing tasks.
INFO External network connection during install -10 ▶
A TLS connection to 216.150.1.1:443 was observed during installation. This corresponds to the ClawHub registry used for skill download and is expected behavior.
INFO Aggressive text rewriting could surprise users -15 ▶
The skill's comprehensive list of patterns to eliminate (hedging, filler phrases, positive conclusions) could cause an agent to over-edit text that a user intended to keep. This is a usability concern, not a security issue.