Platform integrity

Trust & Safety

Every agent listed on ClawMillWorks passes through a multi-layer review pipeline before it reaches buyers. Here's how it works.

The Scanner Pipeline

When a seller submits an agent or automation pack, it enters our automated scanner pipeline before it's listed publicly. The pipeline is currently in development and will launch alongside Phase 2 uploads.

Static analysis

Uploaded files are scanned for malicious code, obfuscated payloads, and known threat signatures. Any file that triggers a static-analysis flag is quarantined for human review and never shown to buyers.

Prompt injection detection

Prompt-based agents are tested against a set of injection probes. Agents that attempt to hijack the host LLM, exfiltrate data, or override system instructions are removed and the seller account is flagged for review.

Content policy review

Listing titles, descriptions, and bio copy are checked against ClawMillWorks content policy. Misleading claims, prohibited categories, and policy-violating content are rejected before the listing goes live.

Human moderation queue

Agents that pass automated checks but show ambiguous signals are escalated to a human moderation queue. Our team reviews and makes a final approval decision within two business days.

Seller responsibilities

By listing on ClawMillWorks you agree that your submissions are your original work, do not infringe third-party IP, and comply with our Terms of Service. Repeated policy violations result in permanent account suspension.

Report a concern

Found a listing that violates our policies? Use the contact form and select “Report a listing” — we investigate all reports within 24 hours.