skill-routing

Paper proposes Skill++, which rewrites only the name and description of incoming agent skills to prevent routing interference in growing skill libraries, raising full-skill selection from 36.4% to 76.3% across 731 skills, three benchmarks, five routers, and the Codex CLI harness.

Microsoft AI introduces TalentRL, a strategy-guided reinforcement learning framework for hierarchical routing over procedural memory in LLM agents, where coarse-grained 'talents' guide selection of fine-grained skills. It improves success rates across seven QA benchmarks and achieves 99.3% on ALFWorld and 90.7 on WebShop across Qwen and Llama backbones.

The paper introduces Gavel, a method that elicits native skill routing from frozen LLMs via linear projections, enabling efficient tool selection without context overload and outperforming existing pipelines on benchmarks.

This paper proposes Diverse Skill Routing (DSR), a diversity-aware reranking framework using Determinantal Point Processes to balance relevance and non-redundancy in skill selection for LLM agents, improving recall and coverage on benchmarks.

Routed is an open-source, local-first hybrid router for AI agent skills that runs offline on CPU using a combination of dense embeddings, BM25, and exact matching to route prompts in under 20ms without API calls. The tool supports major agentic environments including Cursor, Claude Code, and MCP servers, and just released v1.1.0 with multilingual support and an MCP server mode.

This paper presents a case study on skill discovery and routing in a multimodal agent harness, showing that partial in-prompt exposure of skills can create lexical competition that hinders correct selection, linking small-scale in-context retrieval to large-scale approaches.

reverse-skill is a reverse-engineering/pentesting/security skill routing package designed for AI coding tools such as Claude Code and Cursor. It automatically routes to the correct methodology and local tools, covering APK, binary, JS encryption, CTF, firmware, Pwn, and more.

Highlights three recent AI papers: SpatialClaw (training-free spatial reasoning via code), SkillWeaver (compositional skill routing with decompose-retrieve-compose pipeline), and PreAct (compiling agent runs into fast state machines for repeated tasks).

Introduces SkillWeaver, a decompose-retrieve-compose framework for routing multiple skills to LLM agents, along with CompSkillBench, a benchmark of 300 compositional queries over 2,209 real MCP server skills.

This paper argues that agentic AI's next bottleneck is system scaling (designing the 'harness' around foundation models), not just model scaling, and introduces CheetahClaws, a Python-native reference harness, along with an analysis of three core bottlenecks: context governance, trustworthy memory, and dynamic skill routing.

This paper argues that advancing agentic AI requires scaling the system architecture around foundation models, focusing on auditable, modular, and verifiable components. The authors introduce CheetahClaws, a reference harness, and outline bottlenecks in context governance, trustworthy memory, and dynamic skill routing.