The package is an intentional remote AI coding runner with privileged local execution and first-party MCP setup, so it warrants a warning as a dangerous capability. Its activation is explicit and source inspection found no concrete malicious install-time or stealth behavior.
Static reason
One or more suspicious static signals were detected.
Trigger
User runs acc-runner watch after login and a control-plane task is assigned.
Impact
An authorized control-plane task can cause the local agent to modify the configured repository and use its available credentials/tools.
Mechanism
Remote task dispatch to privileged local AI-agent subprocesses, with temporary MCP extension setup.
Evidence
package.jsondist/cli.jsdist/task-runner.jsdist/engines/claude-code.jsdist/mcp-spawn.jsdist/runtime/self-upgrade.jsdist/companion-autostart.js.mcp.json
Network endpoints2
agent-control-center-sepia.vercel.app