← Back to Skills Marketplace
Qwen Orchestrator
by
ciklopentan
· GitHub ↗
· v1.5.8
· MIT-0
243
Downloads
0
Stars
0
Active Installs
12
Versions
Install in OpenClaw
/install qwen-orchestrator
Description
Qwen Chat (chat.qwen.ai) access via Puppeteer browser automation with CDP interceptor. Persistent daemon (~35ms startup), health checks, graceful shutdown, P...
Usage Guidance
This skill is coherent with its stated purpose, but it runs a persistent Chromium instance that uses a userDataDir (.profile) and writes session/endpoint/log files in the skill folder. Before installing: (1) review the code yourself (already included) to ensure no remote exfil endpoints; (2) do not point it at your main browser profile — create an isolated profile directory or use a container/VM; (3) be aware the daemon (pm2) will keep a headless browser process running and can access cookies/localStorage/open tabs in that profile; (4) prefer running with --visible/--wait initially to inspect auth flows; (5) if you need tighter guarantees, run the skill in a sandboxed environment with limited network egress and no sensitive accounts logged in. If you trust the author and will use an isolated profile, the skill appears usable; otherwise treat it cautiously because of persistence and local-profile access.
Capability Analysis
Type: OpenClaw Skill
Name: qwen-orchestrator
Version: 1.5.8
The qwen-orchestrator bundle is a sophisticated Puppeteer-based automation tool designed to provide CLI and agent access to the Qwen Chat web interface. It features a persistent background daemon (managed via PM2), session continuity logic, and automated UI interactions (model selection, thinking mode, and web search toggles). While it utilizes high-privilege capabilities such as browser profile management and shell execution (via ask-qwen.sh and qwen-daemon.js) to locate binaries and manage processes, these actions are strictly aligned with its stated purpose of providing a low-latency LLM interface without an API key. No evidence of data exfiltration, malicious prompt injection, or unauthorized persistence was found.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description (Qwen Chat via Puppeteer) match the code and runtime instructions. The included scripts (ask-puppeteer.js, qwen-daemon.js, ask-qwen.sh, auth-check, diagnostics) implement the stated functionality and the preference/continuity logic described in SKILL.md.
Instruction Scope
SKILL.md instructs running commands from the skill root and to use a persistent daemon, .qwen.json config, and session files. The runtime will read/write files under the skill folder (.profile, .sessions, .daemon-ws-endpoint, .logs, .diagnostics) and evaluate page DOMs via Puppeteer. This is expected for browser automation, but it expands the agent's scope to local filesystem state and any data present in the browser profile.
Install Mechanism
There is no external install script; full source is included. Dependencies are standard Node/Puppeteer packages (package.json & package-lock.json). No downloads from unknown URLs or extract-from-untrusted sources are present in the manifest. Note: running npm install / puppeteer may download Chromium depending on environment.
Credentials
The skill requests no environment variables or external credentials. However, it uses a Chromium userDataDir (.profile) and will run a browser that reuses stored cookies/auth. That grants the skill access to any information present in that profile (open tabs, cookies, localStorage) — expected for a browser automation tool but potentially sensitive if you reuse your regular browser profile.
Persistence & Privilege
The skill supports running a persistent daemon (qwen-daemon) managed via pm2 and writes endpoint/session files to the skill directory. While always:false, the daemon is explicitly intended to be long-lived and will run headless Chromium with a user profile. Persistent background processes with access to a browser profile increase blast radius if compromised or misused.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install qwen-orchestrator - After installation, invoke the skill by name or use
/qwen-orchestrator - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.5.8
Force Qwen3.6-Max-Preview as the preferred default model before each prompt; add preferredModel config override; refresh docs.
v1.5.7
Fix publish hygiene by excluding transient runtime state and review artifacts from the release package.
v1.5.6
Harden Qwen submit path with keyboard fallback when button submit silently stalls.
v1.5.5
Reliability hardening for follow-up continuity: require visible hydrated history before follow-up submit, use exact-URL hard rebind for shell-only restored chats, remove unsafe title-based chat switching, add detection-path diagnostics, and expand continuity stress tests to multi-round follow-ups.
v1.5.4
Add adaptive oversized-prompt submit detection and shorten/retry fallback; fix giant prompt transport via --file; expand mode-matrix coverage.
v1.5.3
v1.5.3: fix follow-up continuity bug by requiring hydrated restored chat state before follow-up submit and binding submit to the composer/send control instead of relying on page-global Enter; add real mode-matrix and follow-up continuity smoke tests; update runtime docs and release checks.
v1.5.2
Hardening release: tightened publish hygiene, added deterministic fatal exit-code signaling for daemon/unavailable and continuity-failure paths, and clarified the public contract plus trusted session-file boundary for weaker runtimes.
v1.5.1
Hardening release: improved follow-up session continuity hot path, added stronger thinking/search state verification, documented and enforced soft rateLimitMs behavior, added bilingual/help clarity, and cleaned publish packaging to exclude local runtime artifacts.
v1.2.3
Fix follow-up chat continuity: restore exact chat URL, rebind restored chat before submit, and prevent silent branch to a new chat in multi-round sessions.
v1.2.2
Fix false thinking-state detection, isolate daemon requests onto dedicated pages, sync diagnostics completion status, and tighten continue-button matching to avoid false sidebar clicks.
v1.2.1
Stabilize runtime modes, daemon lifecycle, close behavior, session handling, and docs after full live audit.
v1.2.0
Live-verified release: added reliable thinking-mode switching, fresh-chat handling, Qwen SSE completion parsing, daemon-path fixes, and long-form answer extraction.
Metadata
Frequently Asked Questions
What is Qwen Orchestrator?
Qwen Chat (chat.qwen.ai) access via Puppeteer browser automation with CDP interceptor. Persistent daemon (~35ms startup), health checks, graceful shutdown, P... It is an AI Agent Skill for Claude Code / OpenClaw, with 243 downloads so far.
How do I install Qwen Orchestrator?
Run "/install qwen-orchestrator" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Qwen Orchestrator free?
Yes, Qwen Orchestrator is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Qwen Orchestrator support?
Qwen Orchestrator is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Qwen Orchestrator?
It is built and maintained by ciklopentan (@ciklopentan); the current version is v1.5.8.
More Skills