Back to skill

Security audit

Node Connect Diagnostics

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent OpenClaw connection troubleshooting guide, but it tells users to approve the latest pending device without verifying that it is the intended device.

Review this skill before use in environments where others can reach the pairing endpoint. If you install it, inspect pending pairing requests and approve only the intended device; do not rely on `--latest` when unexpected or competing requests may exist.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:97
Finding
Unverified Approval of the Latest Device Pairing Request## Vulnerability Details **File Location**: `SKILL.md`, lines 97–103 **Vulnerability Type**: Insecure device authorization workflow **Risk Level**: Medium **Vulnerable Code Snippet**: ```markdown If the app says `pairing required`: - network route and auth worked - approve the pending device ```bash openclaw devices list openclaw devices approve --latest ``` ``` ### Technical Analysis The instructions recommend approving a pending device with `openclaw devices approve --latest`. This selects a request based on arrival order rather than a verified device identifier, fingerprint, or other trusted identity attribute. Although the preceding command lists devices, the Skill does not explicitly require the operator to confirm that the latest request belongs to the intended companion device. Consequently, request timing becomes an implicit authorization criterion. This creates a race condition when an untrusted party can reach the pairing endpoint and submit a competing request. ### Attack Path 1. The operator exposes the gateway pairing endpoint over a LAN, Tailscale route, or public reverse proxy as described elsewhere in the Skill. 2. A legitimate companion device submits a pairing request. 3. An attacker who can reach the endpoint submits another pairing request after the legitimate request. 4. The attacker-controlled request becomes the latest pending request. 5. The operator follows the documented guidance and runs `openclaw devices approve --latest`. 6. The gateway approves the attacker-controlled device without the operator verifying its identity. Exploitation requires the attacker to reach the pairing endpoint and successfully create a pending request during the relevant pairing window. ### Impact Assessment Successful exploitation could grant an unauthorized device the access assigned to an approved OpenClaw node. The exact privileges depend on OpenClaw's device authorization model and gateway co ...[truncated 246 chars]
Remediation
## Remediation Suggestions 1. Remove the blanket recommendation to use `openclaw devices approve --latest`. 2. Require the operator to inspect all pending requests and compare the request's explicit device ID, fingerprint, platform, and displayed verification code with the intended companion device. 3. Approve only the verified request by its immutable identifier, for example with an explicit-ID approval command if OpenClaw supports one. 4. If explicit-ID approval is unavailable, reject or cancel all unknown pending requests and restart pairing within a controlled, short-lived window. 5. For publicly reachable routes, restrict access to trusted networks or authenticated reverse-proxy clients where practical. 6. Update the guidance to warn that request recency is not proof of device identity and that unexpected requests must never be approved.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Static analysis

No suspicious patterns detected.