Harnessing LLM Agents with Skill Programs
Hongjun Liu, Yifei Ming, Shafiq Joty, Chen Zhao · 2026
Read Harnessing LLM Agents with Skill Programs by Hongjun Liu, Yifei Ming, Shafiq Joty, Chen Zhao (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-17734.txt.
Introduction
Hongjun Liu and three colleagues turn agent skills from passive advice into runnable code. Their HASP framework upgrades lessons from past work into Program Functions that fire on failure-prone states. A Program Function can change the next action or inject corrective context, acting as a guardrail inside the agent loop. HASP works at run time, in post-training, or through controlled evolution of a checked skill library. Run-time functions alone lift web-search reasoning by 25% over a multi-loop ReAct agent.
Catalog record
| CREATOR | Hongjun Liu, Yifei Ming, Shafiq Joty, Chen Zhao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 140685 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.17734