SkillRefiner: Offline skill refinement from your agents' existing execution traces
gregd_nlp · x · 2026-10-03
SkillRefiner launches to address a common problem: deployed agents pile up execution traces. The project proposes automatically refining agent skills offline from the rollouts you already have — turning past execution experience into reusable skills without redeployment. Details are laid out in a thread.
More from coding & agent
- Evals show up in job listings: nearly half of 25 PM openings want AI testing skills — every · 2026-10-03
- Dev says GPT 6.1 Sol far less reliable than Opus 5.5 for fixing his plugin — vista8 · 2026-10-03
- Supabase overhauls local dev for AI agents: no Docker, multi-instance per repo, declarative schemas — ycombinator · 2026-10-03
- Modal VM Sandboxes hit GA: demo runs Docker Compose apps, tests and coding agents — charles_irl · 2026-10-03
- Runway MCP + Opus 5.5: agent-driven video generation demo with a 1999 vibe — tlakomy · 2026-10-03
- Indie dev builds rainy Hong Kong delivery game with Sonnet 5.5 on ThreeJS — socialwithaayan · 2026-10-03