ResearchSkillCraft tool composition benchmarkWweb·medium signalXBlueskyLinkedInCopy linkBenchmark for agent skill abstraction and cross-task reuse. Skill saving/reuse reduces token usage by 80%. Success rate correlates with tool composition ability. Code released. arXiv 2603.00718↳ Follow the threadPolicy dependency / Stack layerA replay of 68,266 real Claude Code requests says plain LRU beats the clever KV-cache policiesGitHubPolicy dependency / Stack layerRIPPLE: an edit confined to one prompt-policy segment changes downstream behavior, so replay candidate edits after previously accepted ones before persistingarXiv 2609.12127Policy dependency / Stack layerMOSAIC Picks a GraphRAG Traversal Policy per Query and Beats the Best Fixed Policy by 9.96 PointsarXiv 2609.11065Stack layer / Threat patternSnyk put its agent-skill scanner behind a free web page called Skill InspectorSnyk LabsStack layer / Threat patternPattern: Anthropic's own bar is that Claude-written production code gets reviewed harder than human-written codeSimon WillisonPolicy dependency / Stack layergit-ai is a Git extension that tracks which code in your repo was AI-generated, at 140 open PRs to 71 issuesGitHub TrendingPolicy dependency / Stack layerMicrosoft publishes a 37-page 'humanist AI code of conduct' for its MAI models and opens it to public consultationThe Verge (corroborated by Unite.AI and TechBriefly)Policy dependency / Stack layerCodeBLEU Scored 91% for Both RAG Strategies While One of Them Hallucinated APIs 56.4% of the TimearXiv 2609.12464