Kroki’s Mermaid companion renders diagram images using a headless Chromium engine without a network request policy. From version 0.26.0 through 0.31.2 an attacker who can submit Mermaid source files can cause the engine to perform HTTP or HTTPS GET requests to arbitrary URLs. The resulting SSRF is blind: the companion does not capture response content, so the impact is limited to reachability and potential state‑changing GET endpoints. No data exfiltration, cloud metadata access, or port scanning has been demonstrated. The vulnerability is triggered only when the Mermaid companion, PNG rendering, a diagram‑submission path, and network connectivity from the companion to the target are available.

Affected vendors include yuzutech; product Kroki. All releases from v0.26.0 up to and including v0.31.2 are vulnerable. Versions prior to 0.26.0 or after 0.31.2 contain the fix. The issue is specifically tied to the mermaid/src/worker.js component of the Mermaid companion.

The CVSS score of 5 indicates moderate severity. The EPSS score is not available, and the vulnerability is not listed in the CISA KEV catalog, suggesting no broad exploitation campaigns have been reported yet. Exploitation requires that the attacker meets any authentication or authorization mechanisms configured on the Kroki deployment and that the Mermaid companion has outbound network connectivity. Because the SSRF is blind and limited to GET requests, the range of potential damage is narrower than typical SSRF scenarios, but an attacker could still reach internal services or trigger side‑effects on external state‑changing endpoints.