Clawdrain: Exploiting Tool-Calling Chains for Stealthy Token Exhaustion in OpenClaw Agents DOI: XXXXXXX.XXXXXXX Conference: ; ;
Ben Dong, Hui Feng, Qian Wang · 2026
Read Clawdrain: Exploiting Tool-Calling Chains for Stealthy Token Exhaustion in OpenClaw Agents DOI: XXXXXXX.XXXXXXX Conference: ; ; by Ben Dong, Hui Feng, Qian Wang (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-2603-00902.txt.
Description
Dong, Feng, and Wang expose a token-drain attack on OpenClaw, an open-source self-hosted assistant with a community skill ecosystem. Their trojanized skill steers the agent through a staged verification protocol via injected instructions and a helper script. Token use grows six to seven times over a benign baseline, near nine times in a costly failure setup. The agent sometimes routes around brittle steps with general tools, softening the drain, and the authors map further production attack routes.
Catalog record
| CREATOR | Ben Dong, Hui Feng, Qian Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35980 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.00902