Run Agent Skills in a Desktop Client With Cherry Studio

Agent skills usually live next to a terminal. Cherry Studio puts them inside a desktop chat client, alongside 300+ pre-configured assistants and multi-provider model access — so the skills run where the conversation happens.

Why This Skill Matters

Cherry Studio is a desktop client for Windows, Mac, and Linux that fronts cloud services (OpenAI, Gemini, Anthropic), AI web apps (Claude, Perplexity, Poe), and local models (Ollama, LM Studio) in one interface, per its README. Less known: the repo ships an agent-skills runtime. Five skills sit in resources/skills/ (verified in the repo tree on October 11, 2026), and the bundled find-skills skill lets your assistant search a public marketplace and install new skills mid-conversation. If you want skills without living in a shell, this is the GUI path.

Installation

Download the installer for your platform from the GitHub Releases page. The community edition is AGPL-3.0, and the README describes it as ready to use with no environment setup — there is no build step to run.

Real Workflow: Install a Skill From the Marketplace

The built-in find-skills skill handles discovery. Its SKILL.md hands the agent two tools: search_skills(query) to search the marketplace by keyword, and install_skill(install_source) to install one skill into Cherry's managed library and enable it for the agent. Per the same file, Cherry clones the repo, installs just that one skill, and registers it in a single deterministic step, with permission handling following the active Claude permission mode. Ask in plain language:

Find me a skill that can scrape Google Maps and install it.

The four other bundled skills cover the desktop-specific ground a terminal agent can't: cherry-browser drives the app's visible Agent browser for navigation, forms, screenshots, and authenticated sites; office-transform extracts or converts part of a spreadsheet, Word document, PDF, or slide deck into a new file without touching the source; skill-creator creates new skills and runs evals against them; cherry-tool-guide routes the agent to Cherry's first-party tools — knowledge, memory, schedules, image generation, and more — before it improvises a shell workaround.

Tips

• The marketplace find-skills browses is skills.sh, per its SKILL.md — the same ecosystem the npx skills CLI installs from.
• cherry-browser needs the Browser setting enabled and an available Agent pane, per its own description.
• office-transform always writes a new file; per its description, the source file is never modified.
• The README documents multi-model simultaneous conversations — useful for watching two models use the same skill on the same prompt.
• Cherry Studio also supports MCP servers, so installed skills and MCP tools land in the same client.

When Not to Use This

Cherry Studio is an interactive desktop app — headless or CI automation still belongs to a CLI agent. And its skills runtime is young surface area: the main README documents the client features but not the skills feature, so treat the bundled five as the verified inventory and expect the marketplace flow to evolve.

See the leaderboard for more skills.