SkillClone: Multi-Modal Clone Detection and Clone Propagation Analysis in the Agent Skill Ecosystem

2026-03-23

AI Summary

Current multimodal agent skills—encompassing YAML, natural language, and code—lack mechanisms for clone detection, leading to silent propagation of vulnerabilities and difficulties in provenance tracing. This work proposes the first clone detection method tailored for such multimodal skills, integrating holistic TF-IDF similarity with decomposed modality-specific features and employing logistic regression to achieve high-precision clone identification and interpretable type classification. We introduce SkillClone-Bench, a benchmark comprising 300 annotated skill pairs, on which our approach attains an F1 score of 0.939. A large-scale analysis of 20,000 skills reveals 258,000 clone pairs, with 75% of skills involved in cloning; only 5,642 are truly unique, indicating a redundancy rate of 3.5× and highlighting the severity of cross-author cloning and ecosystem-wide redundancy.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Agent skills are modular instruction packages that combine YAML metadata, natural language instructions, and embedded code, and they have reached 196K publicly available instances, yet no mechanism exists to detect clone relationships among them. This gap creates systemic risks: a vulnerability in a widely copied skill silently persists across derivatives with no alert to maintainers. Existing clone detectors, designed for single-modality source code, cannot handle the multi-modal structure of skills, where clone evidence is distributed across three interleaved content channels. We present SkillClone, the first multi-modal clone detection approach for agent skills. SkillClone fuses flat TF-IDF similarity with per-channel decomposition (YAML, NL, code) through logistic regression, combining strong detection with interpretable type classification. We construct SkillClone-Bench, a balanced benchmark of 300 ground-truth pairs with stratified difficulty. On SkillClone-Bench, SkillClone achieves F1 of 0.939 with precision 0.952, outperforming flat TF-IDF (F1 = 0.881) and achieving 4.2x higher Type-4 (semantic) recall than MinHash. Applying SkillClone to 20K skills reveals 258K clone pairs involving 75% of all skills, with 40% crossing author boundaries. A deduplication analysis shows the ecosystem is inflated 3.5x: only 5,642 unique skill concepts underlie the 20K listed skills, and 41% of skills in clone families are superseded by a strictly better variant.

Problem

• clone detection
• multi-modal
• agent skills
• skill ecosystem
• code clones

Innovation

• multi-modal clone detection
• agent skills
• clone propagation analysis
• skill ecosystem
• interpretable type classification