Paper Detail

ContractSkill: Repairable Contract-Based Skills for Multimodal Web Agents

Zijian Lu, Yiping Zuo, Yupeng Nie, Xin He, Weibei Fan, Chen Dai

arxiv Score 29.4

Published 2026-03-20 · First seen 2026-03-27

Research Track B · General AI

Abstract

Despite rapid progress in multimodal GUI agents, reusable skill acquisition remains difficult because on-demand generated skills often leave action semantics, state assumptions, and success criteria implicit. This makes them brittle to execution errors, hard to verify, and difficult to repair. We present ContractSkill, a framework that converts a draft skill into a contracted executable artifact with explicit preconditions, step specifications, postconditions, recovery rules, and termination checks. This representation enables deterministic verification, step-level fault localization, and minimal patch-based repair, turning skill refinement into localized editing rather than full regeneration. Experiments on VisualWebArena and MiniWoB with GLM-4.6V and Qwen3.5-Plus show that ContractSkill improves self-generated skills from 9.4% and 10.9% to 28.1% and 37.5% on VisualWebArena, and from 66.5% and 60.5% to 77.5% and 81.0% on MiniWoB. Repaired artifacts also transfer across models, improving the target model's self-generated-skill baseline by up to 47.8 points and 12.8 points on the two benchmarks, respectively. These results suggest that agent skills are better treated as explicit procedural artifacts that can be verified, repaired, and shared across models.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
now
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

Key Findings

The ContractSkill framework significantly improves the performance of self-generated skills for multimodal web agents, boosting success rates by over 20 percentage points on both the VisualWebArena and MiniWoB benchmarks. Repaired skills are also shown to be transferable, allowing a skill refined by one model to substantially improve the performance of another. This suggests skills can be treated as shareable, reusable assets.

Limitations

The provided abstract does not explicitly state limitations or directions for future work.

Methodology

ContractSkill converts a draft skill into a structured, contracted artifact with explicit preconditions, postconditions, step specifications, and recovery rules, which enables deterministic verification, fault localization, and minimal patch-based repair.

Significance

This research suggests that treating agent skills as explicit, verifiable, and repairable procedural artifacts is a more robust and scalable approach than relying on brittle, on-demand skill generation.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@article{lu2026contractskill,
  title = {ContractSkill: Repairable Contract-Based Skills for Multimodal Web Agents},
  author = {Zijian Lu and Yiping Zuo and Yupeng Nie and Xin He and Weibei Fan and Chen Dai},
  year = {2026},
  abstract = {Despite rapid progress in multimodal GUI agents, reusable skill acquisition remains difficult because on-demand generated skills often leave action semantics, state assumptions, and success criteria implicit. This makes them brittle to execution errors, hard to verify, and difficult to repair. We present ContractSkill, a framework that converts a draft skill into a contracted executable artifact with explicit preconditions, step specifications, postconditions, recovery rules, and termination che},
  url = {https://arxiv.org/abs/2603.20340},
  keywords = {cs.SE, cs.AI},
  eprint = {2603.20340},
  archiveprefix = {arXiv},
}

Metadata

{}