Skillv2.0.1

ClawScan security

openclaw-pc-security · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignMar 17, 2026, 10:28 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's code, instructions, and resource usage are consistent with a local Windows + OpenClaw security self-check; the active network and credential checks it contains are expected for this purpose and are documented as requiring authorization.
Guidance: This package appears to do what it says: local Windows posture checks and optional OpenClaw network probing. Important things to consider before running: - Only run active scans (--ports, credential checks, exposure checks) against machines and networks you own or have explicit permission to test. The code will attempt default-login POSTs to /login and GETs to unauthenticated endpoints. - Run inside a controlled environment (separate user account or VM) if you are concerned about side effects. - The MSRC API key is optional; only provide it if you want CVE→KB lookups. If you don't provide it that functionality stays disabled. - The scripts will read local config files (cwd config.json, ~/.openclaw/config.json, APPDATA path) to detect server settings — don't run if you don't want those files inspected. - Reports are written to output/ (fixed filenames and may overwrite). Do not upload reports publicly; they can contain sensitive info. - If you want extra assurance, review scripts/run_scan.py and scripts/run_audit.py before executing, and run the audit-only mode first (no network probing) to see results.

Review Dimensions

Purpose & Capability: okThe name/description match what the bundle does: local Windows posture checks, OpenClaw version/config inspection, optional network probing of OpenClaw endpoints, and generation of local HTML/JSON reports. Required binaries/env vars are minimal/optional (requests, optional MSRC API key) and align with the described capabilities.
Instruction Scope: noteThe SKILL.md explicitly instructs running local audit and optional network scans (scripts/run_audit.py and scripts/run_scan.py). The code performs authorized network probes, default-credential login attempts (admin:openclaw), and unauthenticated endpoint retrievals to detect sensitive fields — these are appropriate for an OpenClaw exposure scanner but are active operations with legal/ethical implications. The skill also scans local config paths (cwd config.json, ~/.openclaw/config.json, APPDATA) which aligns with server-config checks. The SKILL.md warns not to use active scans on systems you don't own and not to upload reports.
Install Mechanism: okNo automated install spec is provided (the package is instruction/code-only). Dependencies are limited to 'requests' (requirements.txt). The project expects the user to set up a Python venv and pip-install requirements; nothing in the manifest points to downloads from untrusted hosts or opaque install actions.
Credentials: okNo required environment variables are declared. An MSRC API key is optional and only used for MSRC SUG lookups when the user requests that function. The scripts read local system state (registry, netstat, files under home/CWD/APPDATA) which is necessary for the stated local-audit purpose. No unrelated cloud credentials or broad secrets are requested.
Persistence & Privilege: okThe skill does not request always:true and does not modify other skills. It will run as invoked and does not request permanent elevated platform privileges. Autonomous invocation is allowed by platform defaults but is not combined with other concerning flags.