The OpenClaw ACP bridge does not enforce prompt size limits, allowing an attacker or user to construct an oversized prompt payload that is assembled before being sent to the chat service. This can overload local resources, cause significant performance degradation, and in extreme cases result in a denial‑of‑service for the local ACP client. The weakness is a classic uncontrolled resource consumption flaw (CWE‑400).

The vulnerability affects the OpenClaw personal AI assistant, specifically version 2026.2.17 and earlier, and targets local ACP client integrations such as IDE plugins that communicate over standard input/output streams.

The CVSS score of 4.8 indicates a medium severity, while the EPSS score of less than 1% suggests a low likelihood of exploitation in the near term. The issue is not currently listed in the CISA KEV catalog. Because the vulnerable code path only exists for local clients, exploitation requires an attacker to run a local OpenClaw process or to have a user supply an unusually large prompt. The fix was introduced in the 2026.2.19 release, eliminating the size‑validation gap.