qwen-cloudlisted

# Qwen Cloud Executor Use this skill when Luis asks to use Qwen from Codex Cloud, especially for bounded research, mechanical edits, test writing, docs, or parallel exploration. Qwen is an external executor in Codex Cloud. Codex remains responsible for planning, secrets hygiene, review, verification, and any final file changes. ## Requirements - `~/.local/bin/qwen-agent` exists. - `DASHSCOPE_API_KEY` is available to the agent phase, or stored at `~/.qwen/settings.json` as `{"env":{"DASHSCOPE_API_KEY":"..."}}`. - Default model: `qwen3.7-plus`. - Provider endpoint: `https://dashscope-intl.aliyuncs.com/compatible-mode/v1`. Do not use `BAILIAN_CODING_PLAN_API_KEY` for headless cloud tasks. The local wrapper uses the DashScope Standard OpenAI-compatible endpoint. ## Smoke Check Before claiming the Qwen lane works in a new cloud environment, run: ```bash ~/.local/bin/qwen-agent --dangerously-skip-permissions -p "Reply with exactly: QWEN OK" ``` Expected output: ```text QWEN OK ``` If the key is unavailable, say the Qwen lane is installed but not live-verified. ## Delegation Pattern 1. Write a short acceptance contract: task, allowed files, disallowed actions, verification command, and required output format. 2. Save Qwen output under `.codex/qwen-runs/<timestamp>/` unless the user asked for a different artifact path. 3. Prefer asking Qwen for a patch, analysis, or explicit command transcript. 4. Review Qwen output before applying changes. 5. Run the repo's ve