Skills on the Fly: Test-Time Adaptive Skill Synthesis for LLM Agents
Enables LLM agents to adapt to novel tasks on the fly without retraining, addressing the limitation of static skill libraries for test-time tasks.
SkillTTA synthesizes task-specific textual skills from retrieved training trajectories at test time, improving LLM agent performance without parameter updates. On SpreadsheetBench, Pass@1 rose from 0.397 to 0.505; on BigCodeBench, from 0.517 to 0.651.
LLM agents benefit from reusable skills, yet test-time tasks often require guidance more specific than a static skill library can provide. We propose \emph{SkillTTA}, a Test-Time Adaptive Skill Synthesis method that retrieves a small set of training trajectories relevant to the current task and synthesizes them into a temporary, task-specific textual skill. The solver model is kept fixed, so adaptation happens entirely through generated context rather than parameter updates. We evaluate the method on SpreadsheetBench, ALFWorld, and BigCodeBench. Compared with static trajectory-to-skill synthesis using GPT-5.5, task-specific skills improve SpreadsheetBench Pass@1 from 0.397 to 0.505 and BigCodeBench Pass@1 from 0.517 to 0.651. On ALFWorld, the method matches a heavier memory-learning baseline within four points of success rate while producing the shortest successful trajectories among reported methods. Ablations on SpreadsheetBench further show that synthesized skills outperform raw trajectory prompting, that top-$k$ retrieval should stay small, and that failed trajectories are especially useful because they expose recurring evaluator-facing mistakes.