Executive brief
NanoClaw, a tool used to connect messaging apps like Slack and Discord to AI agents, contains a security flaw in how it handles local approval requests. An attacker with access to the same computer where NanoClaw is running can bypass security checks to approve their own access requests by impersonating an authorized administrator. This could allow unauthorized users or malicious software on the system to gain control over agent groups and messaging channels.
Technical details
An improper authorization vulnerability (CWE-285/CWE-266) exists in the `createChatSdkBridge.setup` function within `src/channels/chat-sdk-bridge.ts`. The application starts a local-only webhook on 127.0.0.1 to handle forwarded interaction events but fails to authenticate the sender or validate the `interaction.user.id` provided in the JSON POST body. A local unprivileged process can forge a POST request containing the ID of a legitimate approver. Because the downstream authorization check in `src/modules/permissions/index.ts` relies on this untrusted input, the attacker can successfully bypass the approval process to wire messaging groups to agent groups. As of the advisory date, the project has not responded to the issue report.
Affected products
- nanocoai NanoClaw up to 2.0.64
Timeline
- 2026-06-14: disclosed: Issue reported via GitHub issue #2761
- 2026-07-26: advisory: CVE published via NVD/VulDB