EvoSkill-GUI: training-free skill evolution lifts GUI agents by up to 16.2%
Bofan Chen · hf · 2026-09-18
EvoSkill-GUI is a training-free framework that treats agent skills as living procedural knowledge revised from execution feedback at deployment time. Each skill is a structured multi-file package (plans, backup localization, failure-recovery rules, failure cases), refined via a reflect-revise-reuse loop: the executor revises in-rollout, an isolated critic diagnoses failed trajectories under strict information isolation, and skill files are edited through a restricted tool interface. Across MobileWorld, AndroidWorld, and OSWorld, it improves multiple base models without training by up to +16.2%, +6.0%, and +10.5%. Code is open source.
More from coding & agent
- DeepSeek V4.1-Flash vs V4-Pro Benchmarked: 3x Cheaper but Slower and Weaker Output — Arindam_1729 · 2026-09-18
- The Agent-Coding Failure Nobody Talks About: Right Code, Wrong Idea — gandazgul · 2026-09-18
- Open-source MCP connector lets you drive the JSON-only Jev model via Claude — purealgo · 2026-09-18
- Junior backend dev asks: is learning RAG, vector DBs and agents worth it yet? — Effective_Emphasis21 · 2026-09-18
- WebMCP draws developer buzz as a way to bring MCP into the browser — nickbaumann_ · 2026-09-18
- Demo: Claude with built-in Cowork and Artifacts builds a multiplayer game from one prompt — felixrieseberg · 2026-09-18