Skip to content
armadillo

opencode-pimp

Active router for ALL OpenCode requests — classifies and routes to the correct opencode-* skill before any response. Use when anything involves OpenCode CLI, configuration, agents, tools, MCP, plugins, SDK, or GitHub integration.

ModelSource
sonnetpack: opencode
Full Reference If the request involves OpenCode in ANY way — configuration, agents, tools, MCP, plugins, commands, skills, CLI, TUI, SDK, server, ACP, GitHub, Zen, providers, models, themes, keybinds, formatters, LSP, permissions, or anything else OpenCode-related — you MUST route through this skill FIRST.

This is not optional. This is not negotiable. You cannot skip this.

OpenCode Pimp

Section titled “OpenCode Pimp”

The orchestration layer for all OpenCode expertise. Not documentation — an active router. Every OpenCode request flows through this routing table before any response.

Mandatory Announcement — FIRST OUTPUT before anything else:

┏━ 🎛 opencode-pimp ━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ [one-line description of what request/routing]  ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

No exceptions. Box frame first, then route.

Quick Context

Section titled “Quick Context”

OpenCode is the open-source AI coding agent (v1.2.10, Feb 2026). Go-based CLI with TUI, 100k+ GitHub stars, 75+ LLM providers, client/server architecture. Created by the SST/terminal.shop team.

Routing Table

Section titled “Routing Table”

Classify the request. Invoke the matching skill. No response before invocation.

Request PatternSkillWhat It Covers
opencode.json, config sources, providers, models, themes, keybinds, formatters, LSP servers, compactionopencode-configConfiguration system
Build/Plan agents, custom agents, subagents, agent modes, agent creationopencode-agentsAgent system
Built-in tools, custom tools, permissions, tool control, deny/allow/askopencode-toolsTools & permissions
MCP servers, plugins, hooks, custom commands, agent skills, rules, AGENTS.mdopencode-extensionsExtensibility layer
CLI commands, TUI, server mode, SDK, ACP, GitHub integration, Zenopencode-cliCLI & interfaces

Cross-Cutting Rules

Section titled “Cross-Cutting Rules”
  • If a request spans multiple skills, invoke the PRIMARY skill first (the one closest to the core question)
  • If unclear which skill fits, route to opencode-config as the default — it covers the broadest surface
  • Migration questions (Claude Code → OpenCode) → opencode-extensions (covers CLAUDE.md compat)
  • “How do I set up OpenCode?” → opencode-cli (installation/setup)
  • “How do I customize OpenCode?” → opencode-config (configuration)
  • “How do I extend OpenCode?” → opencode-extensions (MCP/plugins/skills)

When Multiple Skills Apply

Section titled “When Multiple Skills Apply”

Priority order:

  1. Config — if the question involves opencode.json settings
  2. Agents — if the question is about agent behavior or creation
  3. Tools — if the question is about what the LLM can do
  4. Extensions — if the question is about adding new capabilities
  5. CLI — if the question is about running or interfacing with OpenCode

What This Skill Does NOT Route

Section titled “What This Skill Does NOT Route”
  • General coding questions (even if using OpenCode) → let armadillo-shepherd handle normally
  • Questions about OTHER tools (Claude Code, Cursor) → not OpenCode-specific
  • LLM model questions (how Claude works) → not OpenCode-specific

Hard Rules

Section titled “Hard Rules”
  • Never respond about OpenCode before invoking the target skill
  • No summarizing, planning to invoke, or explaining what you’re about to do
  • If unclear, ask ONE clarifying question, then route
  • The skill’s content has the verified facts — always defer to it