Back to skill
Skillv1.0.5
ClawScan security
gateway-notify · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 9, 2026, 12:55 PM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's code and instructions match its stated purpose (creating a gateway-startup hook that sends notifications); the implementation is self-contained and does not request unrelated credentials, though it invokes external messaging CLIs which you should verify are trusted and installed.
- Guidance
- This skill appears coherent with its purpose. Before installing: (1) inspect scripts/setup_gateway_notify.sh yourself — it creates files under ~/.openclaw/hooks and writes handler.ts; (2) ensure the messaging CLIs it will call (imsg, wacli, openclaw message) are the official/trusted binaries on your system — a malicious replacement binary with one of those names could be invoked; (3) optionally open the generated handler.ts after running the setup script to confirm the embedded address and command look correct; (4) if you are concerned about command injection, test with a dummy address and review the SAFE_ADDRESS escaping in the setup script. No extra credentials are requested by this skill.
Review Dimensions
- Purpose & Capability
- okName and description (notify on gateway startup) align with the files and script. The setup script only creates a hook under ~/.openclaw/hooks/gateway-restart-notify, generates HOOK.md and handler.ts, enables the hook, and asks the user to restart the gateway. No unrelated cloud credentials or global config paths are requested.
- Instruction Scope
- noteSKILL.md and setup script stay within the described scope: they write to the skill's hook directory, enable the hook, and the handler only collects timestamp and a hardcoded port. The handler executes external CLI commands (imsg, wacli, openclaw message) to send notifications — this is expected for a notification skill but means the skill will invoke local binaries. The script performs input validation and attempts to escape single quotes before embedding the address into handler.ts.
- Install Mechanism
- okInstruction-only skill with no install spec; the only code written to disk is created by the provided setup script in the user's home directory. No remote downloads or archive extraction are performed by the skill itself.
- Credentials
- okNo environment variables, secrets, or external credentials are requested. The messaging address is provided by the user at setup. The handler does not read OpenClaw config files (project changelog/SECURITY.md explicitly documents removal of config reads).
- Persistence & Privilege
- okThe skill does not request always: true and does not modify other skills. It enables a hook via the OpenClaw CLI (openclaw hooks enable) which is the expected mechanism for registering event handlers.