EfficientSkill: Learning Skills for Token-Efficient Multi-Turn Agents

Deep Shah, Sanket Badhe and Nehal Kathrotia at Google, with Priyanka Tiwari at Purdue, introduce EfficientSkill, which first learns a skill for accuracy and then prunes it for token use under an accuracy guard, moving reusable code into workspace utilities that cost no prompt tokens.
Ask this paper
Problem. Reflective skill optimizers add edge-case rules after each failure. The skills grow, token use rises, and agents fall into long loops that time out.
Two phases. Phase 1 repairs failed rollouts without any length penalty to reach a validation accuracy A*. Phase 2 starts from that skill and evolves it to reduce output tokens while keeping accuracy close to A*. Starting the pruning from the minimal seed instead stalls on unfamiliar tasks.
Results. Across four models on MATH-500, ALFWorld and SpreadsheetBench, output tokens fall on 11 of 12 configurations compared with EvoSkill (38% to 80% on 9 of them), and mean accuracy rises on every benchmark.
Cost. On SpreadsheetBench, held-out accuracy rises 14.3 points with Gemini 3.8 Flash and 17.5 points with Grok 4.1 Fast. Billed input tokens fall 8% to 66%, and serving cost falls 25% to 67% without caching across all 12 pairs.
Abstract
Reflective skill optimizers improve multi-turn LLM agents by revising instructions from failed and successful rollouts, yet accumulating edge-case rules bloats prompts, inflates token use, and pushes agents into long, error-prone loops. We introduce EfficientSkill, a two-step framework that first discovers a reliable skill without token penalties, and then warm-starts accuracy-guarded pruning from that skill to compress instructions, bound agent turns, and offload reusable code into workspace utilities with zero implementation prompt cost. Across four frontier, mid-tier, and open-weight models on MATH-500, ALFWorld, and SpreadsheetBench, EfficientSkill cuts output tokens over unconstrained EvoSkill on 11 of 12 configurations (by up to 80%) while raising mean accuracy on every benchmark, including +14.3 and +17.5 percentage-point gains on SpreadsheetBench with Gemini 3.8 Flash and Grok 4.1 Fast by replacing stalled loops with compact workflows. By shrinking both generated tokens and replayed history, EfficientSkill lowers total token usage and serving cost across every evaluated configuration, with or without full-prefix KV caching.