From opencode-agent
Delegate tasks to an opencode agent that uses alternative AI models (default: zai-coding-plan/glm-5-turbo). Use this skill whenever the user asks to "run opencode", "launch an opencode agent", "use opencode", or wants to delegate a task to opencode. Also use it when the user asks to run a task with a non-Claude model like GLM, or mentions opencode in any context where they want work done.
How this skill is triggered — by the user, by Claude, or both
Slash command
/opencode-agent:opencode-agentThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Delegate tasks to [opencode](https://opencode.ai), an alternative AI coding agent. This skill runs `opencode run` as a subprocess, letting you hand off work to models like zai-coding-plan/glm-5-turbo without leaving your Claude Code session.
Delegate tasks to opencode, an alternative AI coding agent. This skill runs opencode run as a subprocess, letting you hand off work to models like zai-coding-plan/glm-5-turbo without leaving your Claude Code session.
Run the user's task via opencode run, passing the task as the message argument. The default model is zai-coding-plan/glm-5-turbo. If the user specifies a different model, pass it with the -m flag.
opencode run -m zai-coding-plan/glm-5-turbo "<task description>"
If the user requests a different model, override:
opencode run -m <provider/model> "<task description>"
opencode run --prompt "<system prompt>" "<task description>"
opencode run --agent <agent-name> "<task description>"
Parse the request. Extract the task the user wants opencode to perform. If the user specified a model, note it. Otherwise default to zai-coding-plan/glm-5-turbo.
Run opencode. Execute the task using Bash. Choose the right execution strategy based on the task's expected duration:
Short tasks (file creation, simple edits, quick scripts) — run inline with a 2-minute timeout:
opencode run -m zai-coding-plan/glm-5-turbo "<task>"
Medium tasks (multi-file changes, test generation, refactoring) — use a longer timeout up to 10 minutes:
opencode run -m zai-coding-plan/glm-5-turbo "<task>" # with timeout set to 600000
Long-running tasks (large codebases, complex generation, full project scaffolding) — run in the background so the user isn't blocked:
opencode run -m zai-coding-plan/glm-5-turbo "<task>" # with run_in_background: true
When running in the background, you'll be notified when the task completes. Let the user know the task is running and you'll report back when it finishes.
Always pass -m zai-coding-plan/glm-5-turbo unless the user specified a different model.
Verify the result. After opencode finishes, check that the expected output was produced. If opencode created or modified files, read them to confirm. Report the outcome to the user.
Handle errors. If opencode fails or isn't installed, tell the user clearly. Common issues:
opencode: command not found — opencode isn't installed. Suggest npm install -g opencode or checking their PATH.opencode models to see available options.opencode run -c "<follow-up message>".opencode models.opencode providers to set up credentials.npx claudepluginhub etdofresh/claude-marketplace --plugin opencode-agentDelegates tasks to GitHub Copilot CLI non-interactively with multi-model support (Claude, GPT, Gemini), permission controls, output sharing, and session resume. Useful for Copilot delegation requests, model comparisons, or programmatic runs.
Delegates coding tasks (debug, implement, refactor) to OpenAI Codex CLI via codex exec, skipping the Node companion runtime for faster execution. Codex writes code; Claude verifies.
Leverages OpenAI Codex/GPT models for autonomous code implementation, reviews, and sandboxed task execution. Triggers on 'codex', 'use gpt', 'full-auto' etc.