🛡️ MAL-2026-12488 — tt-help-cli-ycl

Malicious code in tt-help-cli-ycl (npm)

Source: amazon-inspector

The package's tt-help watchdog subcommand starts a long-running agent that POSTs heartbeats to a remote server (default hardcoded to http://117.71.53.99:17301, referenced in src/lib/constants.js and the README/run-explore examples) and executes commands returned in the response. In WatchdogAgent, syncCommands maps entries from the server's commands array into child_process.spawn(command, { shell: true }) via ProcessManager.startCommand, with auto-restart — the remote operator selects arbitrary shell commands to run on the installer's host. Each heartbeat body assembled by _buildHeartbeatBody carries hostname, non-internal IPv4 address, OS platform/release/arch, CPU count, memory, node version, uptime, load average, and the contents of ~/.tt-help.json (server, proxy, browser, userId, tuning fields), POSTed as JSON to ${serverUrl}/api/watchdog/heartbeat at the configured interval (default 15s). In addition, _startUpgradeChecker polls the npm registry every 10 minutes and, when a new version of tt-help-cli-ycl is published, invokes npm install -g tt-help-cli-ycl@latest via child_process.exec and re-spawns, so the publisher can push arbitrary new code to every running agent host without user confirmation. The combination of remote-command execution, ongoing host inventory + stored-config exfiltration to a bare-IP destination, and unattended auto-upgrade constitutes a full command-and-control backdoor rather than a legitimate remote-management tool.