Connect OpenClaw-style runs to VizClaw live rooms from a ClawHub-installable skill. Use when you need quick room creation, JSONL/websocket bridging, trigger and agent event streaming, or safe overview-mode visualization.
Security Analysis
medium confidenceThe skill appears to do what it claims (bridge OpenClaw events to VizClaw) and its requirements are broadly consistent with that purpose, but it will read local event streams and transmit them to an external service and the README recommends running code from an external URL — so review and be careful about streaming sensitive data before installing.
Name/description (bridge OpenClaw runs to VizClaw rooms) matches the delivered artifacts: an event-bridging Python script that reads JSONL/stdin/local websocket gateways and POSTs/WSes events to VizClaw endpoints. No unrelated credentials or binaries are required.
SKILL.md and the script instruct the agent/user to tail local OpenClaw JSONL logs, connect to local websocket gateways (e.g. ws://127.0.0.1:18789), or accept stdin piping; those inputs can contain arbitrary text (including secrets). The README does include a short safety note, but the instructions do permit transmitting local agent activity to an external service, so users must avoid piping sensitive data. The scope is consistent with the stated purpose but has privacy/exfiltration risk.
The skill bundle contains the script (no install spec), which is low-risk. However SKILL.md encourages running the script directly from https://vizclaw.com via uv run — that is remote code execution from a third-party domain. Running the remote URL is convenient but increases risk; prefer installing the included code and inspecting it locally before execution.
The skill declares no required env vars or credentials, which is proportional. The script does make outbound HTTP/WS connections to VizClaw and can be pointed at local gateways/files; this is expected, but it means local event data may be sent externally without needing explicit credentials. Users should confirm that event payloads do not contain secrets before enabling streaming.
The skill is not force-included (always:false), does not request system-wide config changes, and contains no install-time persistence mechanisms in the bundle. Autonomous invocation is allowed by default (platform normal) but not escalated by this skill.
Guidance
This skill is coherent with its stated purpose, but it transmits local OpenClaw events to an external VizClaw service. Before installing or running: - Inspect scripts/connect.py locally (the bundle includes the file) rather than blindly running a remote URL. - Avoid piping outputs that may contain secrets (API keys, private prompts, tokens) into the bridge; use overview/hidden modes if available. - If you must test, run it in a restricted environment (network controls or sandbox) and point it at non-sensitive test logs/gateways first. - If you plan to use the uv run https://vizclaw.com URL from SKILL.md, remember that downloads and execution will come from that domain — only do that if you trust vizclaw.com. If you want, I can scan the full connect.py for specific endpoint URLs, payload shapes, and any hardcoded domains or tokens (the provided file is truncated).
Latest Release
v1.0.3
- No visible changes detected in this version. - Functionality and documentation remain the same as the previous release.
More by @araa47
Published by @araa47 on ClawHub
