Evidence · DiggingBeagle record
KnotScryer: security analysis of emergent cross-skill interactions
KnotScryer: security analysis of emergent cross-skill interactions
- Source role
- primary disclosure
Evidence record
KnotScryer: security analysis of emergent cross-skill interactions
Claim-level citations (2)
- supportsKnotScryer finds attack chains formed from individually clean agent skills: KnotScryer reports that 13.7 percent of its detected unsafe compositions were composed entirely of skills another single-skill scanner had classified as safe.
Abstract, 612 insecure compositions and 13.7 percent safe-by-single-skill scanner
- supportsKnotScryer finds attack chains formed from individually clean agent skills: The KnotScryer paper reports 612 unsafe combinations in a 13,598-skill ClawHub evaluation, including chains whose individual components evade single-skill checks. The paper evaluates exploitability in test harnesses, not live victim deployments.
Abstract; threat model for skill composition; evaluation results