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 points over the skill-enabled baseline and 7.8 over No-Skill, with 29.1-point higher execution precision. On BFCL, task success and execution precision improve by 5.7 and 29.5 points, respectively. Zero-shot results on Tau-bench and PopQA suggest partial transfer to unseen domains and skills.

Legal Case Retrieval (LCR) aims to identify cases that are legally similar to a given query case. Existing research has largely recast LCR as the task of modeling the textual similarity between the query case and a historical case. While this approach can handle single-defendant cases fairly easily, such surface-level matching proves insufficient for multi-defendant cases, which typically requires modeling and aligning defendants’ roles and judgment outcomes in different cases. To address the challenges associated with Multi-defendant LCR (Multi-LCR), we (1) present Multi-ELAM and Multi-LeCaRDv2, two multi-defendant datasets we create by enriching widely used benchmarks with defendant-level annotations, and (2) introduce a novel role- and judgment-aware framework for Multi-LCR. Experimental results show that our approach outperforms state-of-the-art baselines, highlighting its effectiveness for Multi-LCR.