openclaw-security-watchdog

OpenClaw 安全巡检工具，一键执行系统安全扫描并生成通俗易懂的报告。 使用场景：用户说"安全巡检"、"安全检查"、"安全审计"、"巡检"、"security audit"、"检查安全"、"系统安全"等。 触发条件：任何与 OpenClaw 安全检测、审计、巡检相关的请求。 运行依赖： 必需：Node.js...

What to consider before installing/running: - Metadata mismatch: The registry metadata claims no runtime binaries required, but SKILL.md and the script require Node.js v18+ and the openclaw CLI plus typical system utilities (find, lsof, ss/netstat, journalctl, etc.). Confirm those are present or the skill will fail. - Data upload is optional but sensitive: The '--push' (完整检测) mode will upload MAC address, hostname, a persistent agent_id stored locally, and the full installed Skill list to https://auth.ctct.cn:10020 (author-operated Changeway service). Only item+brief are claimed to be uploaded, and full details are stored locally, but you must trust the remote service and its operator before using --push. Do not enable --push unless you vet the endpoint/operator. - Persistence & scheduling: The skill will write files under ~/.openclaw (reports, agent_id, baselines) and recommends (and instructs) adding a cron job via 'openclaw cron add'. If you set up scheduling, follow the SKILL.md admonition to never include --push in scheduled jobs to avoid long-term device reporting. - Signature/auth model: The SKILL.md describes an X-SIGN header that is an unhashed/unsigned digest of device identifiers and time (no secret key). That provides weak server-side integrity/anti-replay guarantees; it is not an authentication token. Be cautious about trusting that mechanism. - Practical steps: (1) Inspect the bundled script yourself or run it in a disposable VM/container first. (2) Run in default local/offline mode before considering any upload. (3) Verify the domain owner (auth.ctct.cn) and decide whether you trust that operator to receive device IDs and component inventories. (4) If you add scheduled runs, ensure cron jobs are reviewed and that '--push' is never included. Given these points, the package is not obviously malicious but has privacy and metadata-coherence issues; proceed only after reviewing and, if needed, testing in an isolated environment.

Type: OpenClaw Skill Name: openclaw-security-watchdog Version: 1.1.1 The bundle is a legitimate security auditing tool designed to perform system-wide health checks, monitor sensitive file changes, and audit agent behavior. While it possesses high-risk capabilities—such as scanning network ports, reading system logs, and uploading device identifiers (MAC address, hostname) and a list of installed skills to a remote threat intelligence service (auth.ctct.cn)—these actions are core to its stated purpose. The tool demonstrates a high level of transparency: SKILL.md requires explicit user consent ('2 已了解') before enabling data uploads, and the script openclaw-hybrid-audit-changeway.js specifically redacts sensitive data and excludes technical command details from the uploaded payload. Furthermore, the documentation in references/cron-setup.md explicitly warns against using the upload feature in automated tasks to protect user privacy.