Harnessing LLM Agents with Skill Programs

2026-05-17

AI Summary

Current large language model (LLM) agents primarily rely on static textual suggestions for skill reuse, lacking the capacity for active intervention during execution. This work proposes HASP, a novel framework that, for the first time, translates experiential skills into executable program functions (PFs), enabling dynamic behavioral correction through a modular intervention mechanism when agents encounter error-prone states. HASP supports runtime intervention, post-training supervision, and autonomous evolution of its skill repository. Experimental results demonstrate that HASP significantly outperforms existing approaches across web search, mathematical reasoning, and programming tasks, achieving a 25% performance gain through runtime intervention alone and a further improvement of 30.4% when combined with post-training optimization.

Abstract

Equipping LLM agents with reusable skills derived from past experience has become a popular and successful approach for tackling complex and long-horizon tasks. However, such lessons are often encoded as textual guidance that remains largely advisory, lacking explicit mechanisms for when and how to intervene in the agent loop. To bridge the gap, we introduce HASP(Harnessing LLM Agents with Skill Programs), a new framework that upgrades skills into executable Program Functions (PFs). Rather than offering passive advice, PFs act as executable guardrails that activate on failure-prone states and modify the next action or inject corrective context. HASP is highly modular: it can be applied at inference time for direct agent-loop intervention, during post-training to provide structured supervision, or for self-improvement by evolving validated, teacher-reviewed PFs. Empirically, HASP drives substantial gains compared to both training-free and training-based methods on web-search, math reasoning, and coding tasks. For example, on web-search reasoning, inference-time PFs alone improve the average performance by 25% compared to (multi-loop) ReAct Agent, while post-training and controlled evolution achieve a 30.4% gain over Search-R1. To provide deeper insights into HASP, our mechanism analysis reveals how PFs trigger and intervene, how skills are internalized, and the requirement for stable skill library evolution.

Problem

• LLM agents
• reusable skills
• executable intervention
• agent loop
• skill formalization

Innovation

• Skill Programs
• Program Functions
• LLM Agents
• Modular Intervention
• Self-improvement