qwen-cli-runtime

Qwen Runtime

Use this skill only inside the qwen:qwen-rescue subagent.

Primary helper:

• node "${CLAUDE_PLUGIN_ROOT}/scripts/qwen-companion.mjs" task "<raw arguments>"

Execution rules:

• The rescue subagent is a forwarder, not an orchestrator. Its only job is to invoke task once and return that stdout unchanged.
• Prefer the helper over hand-rolled git, direct qwen CLI strings, or any other Bash activity.
• Do not call setup, status, result, or cancel from qwen:qwen-rescue.
• Use task for every rescue request, including diagnosis, planning, research, and explicit fix requests.
• You may use the qwen-prompting skill to rewrite the user's request into a tighter Qwen prompt before the single task call.
• That prompt drafting is the only Claude-side work allowed. Do not inspect the repo, solve the task yourself, or add independent analysis outside the forwarded prompt text.
• Leave --model unset by default. Add --model only when the user explicitly asks for one.
• Map plus → qwen3.5-plus, max → qwen3-max, turbo → qwen3-turbo, coder → qwen3-coder-plus. The companion does this translation for you, so passing the alias through is fine.

Command selection:

• Use exactly one task invocation per rescue handoff.
• If the forwarded request includes --model, pass it through to task.
• Default to a write-capable Qwen run by adding --write unless the user explicitly asks for read-only behavior.

Safety rules:

• Preserve the user's task text as-is apart from stripping routing flags.
• Do not inspect the repository, read files, grep, monitor progress, poll status, fetch results, cancel jobs, summarize output, or do any follow-up work of your own.
• Return the stdout of the task command exactly as-is.
• If the Bash call fails or Qwen cannot be invoked, return nothing.

Limitations of v0.1:

• Background (--background) execution is not supported yet. Always run foreground.
• --resume/--resume-last is not supported yet. Every rescue starts a fresh Qwen session.
• There is no review or adversarial-review command in v0.1. If the user asks for a review, forward the review request as a plain task prompt that describes what to inspect.