different-ai/openwork
6 skills3.6k installs95.3k starsGitHub
Install
npx skills add https://github.com/different-ai/openworkSkills in this repo
1Opencode PrimitivesOpenCode Primitives is a reference agent skill for builders implementing OpenWork directly on OpenCode—skills, plugins, MCP servers, tools, and configuration-driven permissions. Solo developers using Claude Code, Cursor, or Codex install it when they need authoritative answers about where files live, how discovery works, and which doc pages govern each primitive, so UI copy and automation stay consistent with upstream OpenCode behavior. The skill lists must-cite documentation URLs and applies key facts: SKILL.md YAML frontmatter and naming constraints, walking the git worktree to load skills from .opencode or .claude paths, local versus global plugin directories, and npm plugin listing with Bun at startup. It is intermediate complexity because misconfigured permissions or mismatched skill names silently break agent workflows. Use during Build while wiring agent tooling, and again in Ship review when auditing agent permissions before production enablement.733installs2Opencode Bridgeopencode-bridge is an agent skill for solo and indie builders who ship OpenWork on top of OpenCode. It explains how the UI talks to the runtime through CLI subprocess calls, direct reads of OpenCode’s SQLite store, and an MCP bridge for streaming and permission flows. You use it while implementing the glue layer: spawning `opencode` with JSON output, querying session and message tables, and choosing the right channel when latency or interactivity matters. The readme is reference-style integration knowledge rather than a multi-step ritual, so it fits the build phase when you already chose OpenWork and OpenCode as your stack. It does not replace OpenCode’s own install docs or MCP server authoring; it tells your agent exactly which flags, paths, and tables matter so implementation stays consistent across sessions.630installs3Solidjs Patternssolidjs-patterns is an agent skill that encodes SolidJS reactivity and UI-state conventions used in OpenWork. Solo and indie builders shipping agent or desktop-style UIs often blame “stuck screens” on missing rerenders, but Solid updates through signals—the real failures are coupling bugs, especially one global busy() that disables unrelated controls while a long task waits on a permission answer. This skill teaches fine-grained signals, per-action pending scopes, memo-derived state, and immutable signal writes so overlapping installs, refreshes, and permission dialogs can run safely. Use it when you are building or debugging OpenWork-like Solid interfaces in Claude Code, Cursor, or Codex, before you refactor handlers or add new async flows. It does not replace Solid docs; it gives opinionated patterns that prevent the classic busy/permission deadlock and keep UI state predictable as features grow.571installs4Opencode MirrorThe opencode-mirror skill teaches maintainers of OpenWork-style repos how to keep a fresh local clone of the OpenCode project for self-reference while keeping that tree out of version control. Solo builders automating agent workflows often need readable upstream source beside their own code without submodule ceremony or accidental commits of third-party history. Use it when onboarding a machine, refreshing the mirror before a feature spike, or when an agent needs to cite OpenCode internals from disk. The procedure is intentionally minimal: clone once into vendor/opencode if missing, then fast-forward only pulls on updates. Gotchas are explicit—gitignore the path and refuse merge commits in the mirror so your main repo stays clean. This is operational glue, not a tutorial on OpenCode itself; pair it with your project’s agent skills once the mirror exists. Intermediate complexity reflects git discipline rather than application coding.565installs5Openwork Coreopenwork-core is context and guardrails for contributors working on OpenWork, a native Tauri application that wraps OpenCode for less technical users while still exposing plans, permissions, and progress. Before changing behavior, the skill insists you read the product and architecture docs and keep vendor/opencode mirrored for self-reference. It documents everyday workflows—pnpm tauri dev for desktop, platform-specific iOS and Android dev commands, and how to spawn the OpenCode CLI with JSON output—plus how to read the local SQLite database tables for sessions, messages, and file history. Critical constraints prevent scope creep: do not invent capabilities beyond OpenCode’s tool surface, route filesystem and shell access through Tauri commands, and preserve fluid, mobile-first UI. Solo builders maintaining this repo use the skill across Build when wiring integrations and Ship when validating desktop and mobile builds, because platform setup and threading rules span both. Pair it with the tauri-solidjs skill for UI stack depth.546installs6Tauri Solidjstauri-solidjs is an agent skill for the OpenWork stack combining Tauri 2.x with SolidJS and Vite in a pnpm monorepo. Solo builders use it when they want a single codebase for desktop and optional iOS/Android targets without adopting Electron. The skill encodes the expected folder layout—Rust entry and commands under src-tauri, Solid root in App.tsx, stores for state, and an OpenCode bridge in lib—plus the exact CLI flows for local dev and release builds. It assumes the OpenWork template choices are already aligned so agents do not invent alternate frameworks. Invoke it when bootstrapping the desktop package, wiring Tauri capabilities, or extending UI stores while keeping commands in lib.rs consistent with frontend calls.516installs