OpenClaw EUVD-2026-87474

Exploitation requires the attacker to control or compromise a video-asset provider that the OpenClaw host trusts and queries, so the CLI receives attacker-chosen URLs in provider responses; OpenClaw before 2026.8.1 fails to validate those returned video asset URLs. It further depends on the video-fetch code path being exercised (UI:P/UI:R - a user or agent action that triggers a video asset fetch) and on an internal/loopback service being reachable from the OpenClaw host to have any impact. Limiting factors: the attacker cannot directly reach OpenClaw (impact is mediated entirely through a trusted third-party provider), impact is confined to limited internal requests/disclosure (all impact metrics Low), and versions 2026.8.1+ are patched.

All signals point to a low-priority issue despite the network attack vector. The vendor CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:P/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N) yields only 5.3 (medium), and every impact metric is Low - this is a bounded SSRF that lets a malicious/compromised upstream video provider steer the OpenClaw CLI into requests against loopback/private internal services, not a full read/write compromise. No EPSS score, no CISA KEV listing, and no public exploit are provided, so exploitation status is: no public exploit identified at time of analysis. The critical precondition - the attacker must already control or have compromised a video-asset provider that the host trusts and queries - is a meaningful barrier that the vendor's AC:L/AT:N arguably understates; a realistic 3.1 modeling treats this as higher complexity. Net: a real defense-in-depth gap worth patching on the normal cycle, but not an urgent, widely-exploitable priority. No conflicting signals between sources; the main caveat is the unmodeled trust prerequisite on the provider side.

An attacker who controls or has compromised a video-asset provider that OpenClaw integrates with returns a crafted response containing a loopback or RFC1918 URL (e.g. http://169.254.169.254/ or http://127.0.0.1:port). When a user triggers a video fetch, the CLI dereferences the attacker-supplied URL without validation, issuing a server-side request to an internal service reachable from the OpenClaw host and returning limited response data. No public exploit code is identified at time of analysis.