other
Note
- Location
- SKILL.md:19
- Finding
- Broad Host Environment Reconnaissance## Vulnerability Details **File Location**: `SKILL.md`, lines 19-22 **Vulnerability Type**: Broad environment reconnaissance **Risk Level**: Low ```markdown - **OpenClaw Security Audit:** Runs `openclaw security audit --deep` to find vulnerabilities in your OpenClaw setup. - **Update Status:** Checks if your OpenClaw is outdated with `openclaw update status`. - **Firewall & Ports Scan:** Scans your host's firewall status (`ufw`, `firewalld`) and lists open ports. - **System Vitals:** Basic check of OS version and running services. ``` ### Technical Analysis The skill directs the agent to perform a deep OpenClaw security audit and enumerate firewall status, listening ports, operating-system information, update status, and running services. Although these operations are explicitly presented as read-only and are consistent with the stated health-audit purpose, their combined output creates a detailed profile of the host and its attack surface. The reviewed file contains no instruction to transmit the collected information externally, execute a remote payload, modify the system, obtain elevated privileges, establish persistence, or bypass security controls. The risk is therefore limited to unnecessary collection or disclosure of host metadata through the resulting agent report. ### Attack Path 1. A user asks the agent to run the documented health check. 2. The agent follows the skill instructions and invokes the deep audit and host-enumeration operations. 3. Firewall configuration, open ports, software status, OS details, and running-service information are collected. 4. These details are incorporated into an agent-visible report. 5. If the report is exposed to an unauthorized recipient, that recipient could use the service and port inventory to prioritize further attacks. No automatic external disclosure mechanism or direct exploitation step was identified in the reviewed project. ### Impact Assessment The behavior ...[truncated 493 chars]
- Remediation
- ## Remediation Suggestions - Obtain explicit user consent before inspecting firewall state, open ports, OS details, or running services. - Display the exact commands and categories of data to be collected before execution. - Apply least-collection principles by gathering only information required to calculate the health score. - Redact IP addresses, usernames, service command-line arguments, internal hostnames, and potentially sensitive configuration values from reports. - Store and process audit results locally by default, and require separate confirmation before sharing or transmitting them. - Clearly document the report's retention policy and intended recipients. - Run the audit with the minimum permissions required and avoid privilege elevation unless separately justified and approved.