clone-farm-detector_skill

Overview

This skill detects clone farming and reputation gaming in AI agent marketplaces by analyzing skill content, publishing behavior, and cross-references. It finds near-duplicate skills, ID-washing techniques, batch-publish bursts, and coordinated citation rings. The goal is to surface suspect clusters so marketplaces can prioritize human review and protect search quality and trust.

How this skill works

The detector compares Capsule/Gene code and metadata to compute similarity scores and cluster near-identical skills. It analyzes publishing timelines and node IDs to identify batch patterns and sequential naming. It inspects SHA-256 hashes vs. functional similarity to reveal ID washing and traces dependency and citation graphs to expose coordinated cross-citation. The output is a structured report with clusters, evidence, and a risk rating.

When to use it

• Scan top search results for a competitive keyword to check for clone saturation
• Audit a publisher node to detect mass-publishing or templated uploads
• Monitor marketplace imports or archives for sudden spikes in near-duplicates
• Prioritize candidate skills for manual moderation review
• Assess marketplace health and measure prevalence of reputation gaming

Best practices

• Combine automated flags with human review to avoid false positives from legitimate forks
• Use similarity thresholds tuned to your catalog size and language diversity
• Correlate publication timestamps and node IDs before assigning high-risk labels
• Log evidence (diffs, metadata templates, citation graphs) to support moderation actions
• Periodically re-run scans on top search terms and new publisher activity

Example use cases

• Scan the top 50 results for "code formatter" and produce cluster reports with recommendations
• Audit a suspected publisher node to reveal a batch of 20 near-identical skills uploaded in one hour
• Detect ID washing where multiple unique hashes map to one functional implementation
• Identify cross-citation rings that inflate perceived reputation among a set of skills
• Generate a marketplace health metric showing percentage of suspected clones in popular queries