Skill or Skip? Learning Selective Skill Invocation in Agentic Tasks via Dual-Granularity Preference Learning

2026-05-30

AI Summary

Existing agent skill invocation methods often disrupt decision-making by blindly calling irrelevant skills. This work proposes SelSkill, a novel framework that formulates skill invocation as a selective “call-or-skip” decision problem. By leveraging prediction uncertainty to identify critical decision points and constructing controlled preference pairs based on shared trajectory prefixes, SelSkill jointly optimizes at both the trajectory level (task outcomes) and the step level (invocation preferences). This dual-granularity optimization significantly enhances invocation precision and overall task performance. Empirical results show absolute improvements of 10.9 and 5.7 percentage points in task success rates on ALFWorld and BFCL, respectively, along with over 29% gains in execution accuracy. Furthermore, the framework demonstrates strong zero-shot cross-domain transferability on Tau-bench and PopQA.

Abstract

Agent skills are callable procedural modules that provide reusable knowledge and execution policies for complex agentic tasks. However, existing methods mainly focus on selecting relevant skills or improving the skills themselves, while overlooking whether a relevant skill should actually be invoked at the current decision point. Unhelpful invocations may introduce irrelevant context and disrupt an otherwise correct execution process. To address this issue, we propose SelSkill, a dual-granularity preference-learning framework for selective skill invocation. SelSkill formulates skill use as a skill-or-skip decision, uses predictive uncertainty to prioritize candidate decision points, and constructs controlled invoke-skip preference pairs from shared trajectory prefixes. It further combines episode-level outcome preferences with step-level invocation preferences to capture both overall trajectory quality and the local effectiveness of skill invocation. On ALFWorld with Qwen3-8B, SelSkill improves task success by 10.9 percentage points and execution precision by 29.1 percentage points. On BFCL, it improves task success by 5.7 percentage points and execution precision by 29.5 percentage points. Zero-shot results on Tau-bench and PopQA further suggest that the learned invocation policy transfers to new domains with previously unseen skills.

Problem

• skill invocation
• agentic tasks
• selective execution
• decision-making
• preference learning

Innovation

• selective skill invocation
• dual-granularity preference learning
• skill-or-skip decision
• predictive uncertainty
• controlled preference pairs