Browse agent skills

Let a text-only opencode main agent look at images by routing them to a vision model and working from its text output.

Opencode-style web search for DeepSeek Harness: direct Exa / Parallel MCP calls, no API key required. Registers one ctx.web search provider.

Local GLiNER2.5 context pruning for OpenCode

Adaptive Multi-Agent Orchestration with Cost Intelligence for OpenCode V2

Session-tree token usage, live performance, context window and estimated cost for OpenCode 2

OpenCode V2 plugin: xAI SuperGrok web search, X search, Grok Imagine, speech-to-text, and text-to-speech.

Overwrite the OpenCode system prompt per agent — a raw-system hook that replaces the assembled system with the prompt between markers in your agent file.

Client, CLI, and real tool-call hooks for the hosted eval.seanbehan.ca guardrails service (OpenCode, Hermes Agent, DeepSeek Harness).

OpenCode plugin that installs the optimize-micode-models skill and its /optimize-micode command for intent-driven, quota- and cost-aware model assignments in micode.json; fully dynamic and researched live each run.

Continuous verbatim context pruning for OpenCode, powered by TypeSafe Jev. Port of fast-jev-compaction adapted to OpenCode's context hook.

Local-first context engine for AI coding agents: three persistent graphs (code, project memory, trial-and-error knowledge), deterministic retrieval, native OpenCode plugin. No LLM, no embeddings, no network.

OpenCode plugin and localhost proxy for Qwen3.8 reasoning levels through Ollama

Fast reflex layer (System 1) for coding agents. OpenCode plugin: stuck detection, progress scoring, tool risk policies, completion and context relevance — deterministic first, no LLM in the loop.

Hands-free voice conversation mode for opencode: dictate prompts, answer questions/permissions by voice, with an animated status indicator.

Live tokens-per-second meter for the OpenCode TUI prompt.

MemoryHub memory plugin for opencode — governed, versioned, graph-aware agent memory with semantic search and multi-scope hierarchy.

OpenCode executor: OpenCode CLI via @opencode-ai/sdk/v2.

Devin Fusion multi-agent harness for OpenCode with sandboxed Main and Sidekick agents

Serve the live opencode Go model catalog from the harness's own built-in opencode-go provider route.

Annotate webpages in Chrome and send screenshots + element context to your OpenCode session. Fork of JodusNodus/opencode-chrome-annotation.

One-click local AI installer: OpenCode, Ollama, dynamic model, scrape demo. Linux, Windows, and macOS.

Barada Code – Terminal Coding Agent im OpenCode-Stil. Multi-Provider, interaktiv, gratis startbar.

OpenCode plugin that adds a 'local' provider with auto-detection for Ollama, LMStudio, llama.cpp, vLLM, Exo, llama-swap, oMLX, MLX-VLM, and other OpenAI-compatible servers, with optional per-server API keys

Capability metadata for hand-declared gateway models, resolved at runtime by bare model name (context window, output cap, reasoning levels, vision) — matched against the official DeepSeek route's catalog, the installed pi-ai catalog, opencode and models.d

OpenCode plugin for gbuild — design a feature with the human, plan it as a dependency graph, run it with concurrent fan-out, per-node review, and durable checkpoints.

Reviews OpenCode V2 ask permissions with an LLM before a prompt is shown

Learn any language organically while prompting AI coding agents.

OpenCode V2 plugin: per-session prompt cache hit rate in the footer, sidebar breakdown (tokens, cost, sub-agent rollup), palette command, and cache_hit tool.

Local voice input plugin for OpenCode (V2 TUI port)

Subscription-aware model routing for Pi, OpenCode and every OpenAI-compatible coding agent