---
title: "Generative Skill Composition for LLM Agents"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2606.32025v1"
permalink: "https://feed.astralor.com/items/itm_f30b61d5534941f8a3cac8b3"
published: "2026-06-30T17:53:09.000Z"
score: 59
channel: "ai"
tags: ["LLM agent","SkillComposer","Skill Composition","Structured Planning","Autoregressive Decoding"]
---

# Generative Skill Composition for LLM Agents

## At a glance

LLM agents benefit from skill libraries for complex tasks, but selecting the right skill composition—subset, count, and order—remains a bottleneck. Existing methods either expose all skills or use retrieval, missing the structural nature. This paper formalizes structured skill composition and proposes SkillComposer, a constrained autoregressive decoder that jointly predicts the skill plan. Using a human-curated skill library for training, SkillComposer outperforms retrieval baselines in composition quality and downstream task success, matching the gold-skill upper bound with lower prompt-token cost.

## Key points

- Formalizes structured skill composition: jointly predict subset, count, and order given a task and skill library
- Proposes SkillComposer, a constrained autoregressive decoder that generates skill plans in a single pass, capturing dependencies between consecutive skills
- Builds training set from a real human-curated skill library; evaluations show significant improvements in composition quality and downstream task success
- On GPT-5.2-Codex and Gemini-3-Pro-Preview, SkillComposer boosts pass rates by +23.1 and +18.2 percentage points, matching gold-skill upper bound with lower prompt token cost

## Full text

Recent LLM agents benefit from skills for solving complex tasks. Skills encapsulate modular packages of procedural knowledge and instructions for performing specialized tasks, such as setting up a sandboxed environment, running a test suite, or refactoring a function across multiple files. As skill libraries grow and become reusable across tasks and domains, selecting an appropriate skill composition has emerged as a central bottleneck. Existing approaches fall into two categories. One exposes the agent's reasoning to the entire skill collection; the other performs skill retrieval via embeddings or LLM-based rerankers. Both provide useful insights; however, they miss the structural nature of skill composition, which is a joint decision over which skills, how many, and in what order -- three dimensions that cannot be decoupled. We formalize this as structured skill composition: given a task and a skill library, predict an executable skill plan that jointly specifies the activated subset, count, and execution order. We propose SkillComposer, which instantiates structured skill composition as task-conditioned skill sequence prediction. SkillComposer uses a constrained autoregressive decoder over skill identifiers, so subset, count, and order emerge jointly from a single decoding pass, and dependencies between successive skills are captured naturally. We build a training set of task-composition pairs from a real, human-curated skill library. We then evaluate SkillComposer along two axes: composition quality on a held-out test set, and downstream task success on SkillsBench across two production-grade coding agents. On GPT-5.2-Codex, Gemini-3-Pro-Preview, SkillComposer raises the pass rate by +23.1, +18.2pp over the no-skill baseline, surpassing top-3 retrieval and matching the gold-skill retrieval upper bound at lower prompt-token cost.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_f30b61d5534941f8a3cac8b3