MalwarePublished 6 Aug 2026

@xiaohhhh1/canvas-agent (npm) · Malicious Code

Affects @xiaohhhh1/canvas-agentSource GCVE Database

Malicious package : remove @xiaohhhh1/canvas-agent and rotate any secrets on hosts that installed it.

Data as of 7 Aug 2026, 15:17 UTC

Classification

Malicious package

Published to npm · OSV MAL-2026-13398

What to do

Remove @xiaohhhh1/canvas-agent from every project, lockfile and image that pulls it in. Treat any host that installed it as compromised: rotate the credentials, tokens and keys it could read, from a clean machine.

Affected software

Product	Affected	Fixed in
npm@xiaohhhh1/canvas-agent	Affected =0.4.4	Fixed in–

Check your own hosts for these packages with threat-finder (opens in a new tab), our free open-source scanner.

Description

--- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (a34add291ef7584aec7041449cb2e80d36ae98f93c2725e1aa018707e3334119) On bin start, @xiaohhhh1/canvas-agent opens an outbound WebSocket to the hardcoded relay wss://canvas.xiaohhhh1.com/api/agent-relay and treats inbound messages as HTTP requests to its own loopback API at 127.0.0.1:17371. The bridge unconditionally injects the local agent token (x-canvas-agent-token) into forwarded requests, so remote 'request' messages satisfy token authentication and can invoke /agent/codex/turn and /agent/claude/turn, which spawn the Codex app-server and Claude CLI locally with attacker-supplied prompts and a caller-controlled permissionMode. Codex/Claude are autonomous coding agents that can write files in the workspace and, when permissionMode is widened to danger-full-access, execute arbitrary shell commands. The same code path also bypasses the loopback server's browser-Origin allow-list: the origin gate is populated only after token validation, and the relay bridge presents the valid token without any user-consented browser Origin, so remote-originated calls pass the token+origin check the README advertises. The net effect is a network-to-local-agent-to-shell control path from a hardcoded author-controlled server into the installer's machine and any repository it is run against.

Impact

A remote attacker controlling the relay server can execute arbitrary commands, read arbitrary image files, and open the file manager on the host system running the affected package version. This leads to full remote code execution and unauthorized file access without any local user approval, posing a critical security risk to the host environment.

Defensive guidance

Users should immediately remove the @xiaohhhh1/canvas-agent package version 0.4.4 from their environments and avoid installing it. Since this is a malicious package, the best mitigation is to avoid using it entirely and verify package integrity and provenance before installation.

Summary

The @xiaohhhh1/canvas-agent npm package (version 0.4.4) includes malicious functionality where, upon starting its binary, it opens a WebSocket connection to wss://canvas.xiaohhhh1.com/api/agent-relay. Through this relay, inbound messages are forwarded as authenticated HTTP requests to the local server with a special token header. The relay exposes endpoints that allow spawning local Codex and Claude CLI processes, reading arbitrary image files from absolute paths, and opening the OS file manager on arbitrary absolute paths. The package maps a 'full' permission mode to a 'danger-full-access' sandbox with an approval policy set to 'never', enabling remote peers to execute commands and modify files on the host without any local user prompt or approval. This effectively provides a remote attacker with unrestricted control over the host system via the relay.

References

Community reviews

Loading reviews…

What to do

Affected software

Description

AI analysisMachine-generated, updated 6 Aug 2026, 18:42 UTC

Technical details

References

Community reviews