Skill

opencode-runtime

Internal helper contract for calling the opencode-runner from Claude Code subagents

Install

Run in your terminal

npx claudepluginhub apoapps/swarm-code-plugin --plugin opencode

Tool Access

This skill uses the workspace's default tool permissions.

Skill Content

Similar Skills

cache-components

Guides Next.js Cache Components and Partial Prerendering (PPR) with cacheComponents enabled. Implements 'use cache', cacheLife(), cacheTag(), revalidateTag(), static/dynamic optimization, and cache debugging.

cache-components

138.6k

claude-opus-4-5-migration

2 files

Migrates code, prompts, and API calls from Claude Sonnet 4.0/4.5 or Opus 4.1 to Opus 4.5, updating model strings on Anthropic, AWS, GCP, Azure platforms.

claude-opus-4-5-migration

83.2k

cost-optimization

1 file

Optimizes cloud costs on AWS, Azure, GCP via rightsizing, tagging strategies, reserved instances, spot usage, and spending analysis. Use for expense reduction and governance.

cloud-infrastructure

33.0k

Stats

Parent Repo Stars2

Parent Repo Forks0

Last CommitApr 2, 2026

Actions

View Source View Plugin View on GitHub View README

OpenCode Runtime

Use this skill only inside the opencode:opencode-worker subagent.

Primary helpers:

Ask: node "${CLAUDE_PLUGIN_ROOT}/scripts/opencode-runner.mjs" ask "<prompt>"

Review: node "${CLAUDE_PLUGIN_ROOT}/scripts/opencode-runner.mjs" review [--base <ref>]

Plan: node "${CLAUDE_PLUGIN_ROOT}/scripts/opencode-runner.mjs" plan "<prompt>"

Setup: node "${CLAUDE_PLUGIN_ROOT}/scripts/opencode-runner.mjs" setup --json

Execution rules:

The worker subagent is a forwarder, not an orchestrator. Its only job is to invoke one command and return stdout unchanged.

Prefer the helper script over hand-rolled opencode exec strings or any other Bash activity.

Do not call status, result, or setup from the worker subagent.

Do not inspect the repository, read files, grep, monitor progress, or do follow-up work.

Return the stdout of the command exactly as-is.

If the Bash call fails or OpenCode cannot be invoked, return nothing.

Model control:

Models are auto-detected and configured via /opencode:setup. The runner uses the configured priority list with automatic fallback.

Add --model <model> only when the user explicitly requests a specific model.

The runner script handles fallback to the next available model automatically.

Command selection:

Use exactly one invocation per handoff.

ask for general questions, explanations, debugging help.

review for code review of git changes.

plan for architecture and implementation planning.

Choose the command that best matches the user's intent.