Back to skill
Skillv1.0.0

ClawScan security

Weather Push · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

SuspiciousMar 3, 2026, 4:41 AM
Verdict
suspicious
Confidence
medium
Model
gpt-5-mini
Summary
The skill largely does what it describes (send daily Shenzhen weather and remote service checks to Feishu), but the runtime script contains hard-coded host/user/target values, disables SSH host-key checking, and assumes local binaries/keys without declaring them — these are suspicious and require review before use.
Guidance
This script is coherent with its weather+service-check purpose, but review and adapt it before enabling: - Inspect and change hard-coded values: the SSH target (10.144.1.3), SSH user (sulada2), and the Feishu target OU ID are baked into the script; replace with your own targets or make them configurable. - Provide SSH credentials intentionally: the script expects an SSH key/agent to allow passwordless SSH. Confirm which key will be used and do not expose private keys. - Re-enable host-key verification: StrictHostKeyChecking=no weakens SSH security and can allow MitM; consider using known_hosts or more secure SSH options. - Verify the openclaw CLI path: /home/aisulada/.npm-global/bin/openclaw is user-specific and may not exist; change to the correct path or use a configurable command. - Check logging and confidentiality: logs are written to /tmp/weather-push.log; ensure log files are stored securely if they contain sensitive identifiers. - Test in a safe environment first: run manually and confirm outputs, SSH behavior, and that messages are sent to the expected Feishu recipients. If you want a lower-risk version, request the maintainer remove hard-coded hosts/users, make paths and targets configurable via declared env vars, and avoid disabling SSH host-key checks.

Review Dimensions

Purpose & Capability
okName/description match the behavior: the script fetches weather data, computes differences, checks a remote MiHoMo service, and sends a Feishu message. The declared dependencies (python3 + lunarcalendar, SSH access) are consistent with the stated purpose.
Instruction Scope
concernThe SKILL.md and push.sh instruct the agent to SSH into an internal host (10.144.1.3) as a specific user (sulada2) and run systemctl/pgrep on that host. The script also calls a local OpenClaw CLI at a hard-coded user path and writes logs to /tmp. Disabling StrictHostKeyChecking and hard-coded targets give the script broad, specific network/system access that the user may not expect.
Install Mechanism
okNo install spec — instruction-only plus an included script. Nothing is downloaded or extracted during install, which is low risk. The only runtime requirements are Python packages and SSH access.
Credentials
noteThe skill does not declare env vars or credentials, but it requires SSH credentials/key material to access 10.144.1.3 and expects a user-local openclaw binary path (/home/aisulada/.npm-global/bin/openclaw). These implicit credential assumptions are reasonable for the function but should be explicit; hard-coded username, internal IP, and Feishu target are sensitive and may not be appropriate for other environments.
Persistence & Privilege
okThe skill is not always-enabled and does not modify other skills or system-wide config. It runs as a scheduled/explicit task only and does not request elevated platform privileges.