Verified Humanizer New
Analysis
This skill is a coherent local text-rewriting workflow with privacy guardrails, with only minor notes about optional verification, a bootstrap hook, and packaging metadata consistency.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.
return { role: "system", content: "[Verified Humanizer] Rewrite locally, measure before/after changes, and only verify structured metrics if verification is needed. Never send original or rewritten text to external verification." };The hook can add a system-role instruction when active. The instruction is limited to the skill's stated workflow and privacy constraints, so it is a notice rather than a concern.
"ownerId": "kn71nqqcxyxyst7f2s3f2nzz0h80jf78", "slug": "verified-humanizer", "version": "0.0.7"
The packaged metadata does not match the supplied registry metadata for owner, slug, and version, which lists a different owner ID, slug 'verified-humanizer-new', and version '1.0.0'.
Checks for exposed credentials, poisoned memory or context, unclear communication boundaries, or sensitive data that could leave the user's control.
If verification is used, send only structured evaluation data ... do not include text content ... verification is optional
The skill allows an optional verification step that may send evaluation metrics outside the local rewrite flow, while explicitly forbidding raw original or rewritten text.