Ready workflows for your AI apps
Cheatcode_
Make your AI work like a pro.
Workflows built from how people in each job actually do the work. Find the one you need, or have one laid out for you, and run it in the AI app you already pay for. Your files stay on your computer.
- 1
Find your workflow. Say what you need done. We open the closest ready workflow, or lay out the steps for one that fits your case.
- 2
Open it in your own AI. The free Cheatcode app for Mac sets it up in the AI app you already use and pay for, with your files kept on your computer.
- 3
Your AI works like a pro. It follows the steps a professional would, with the right tools for each, checks its claims and waits for your yes at every stage. Stop anytime and continue later.
01By role
Start from your job. Each role's work as a tree, built from occupational data and the field's own body of knowledge. Editorial, never paid.
The weekly product loop as installable skills: synthesize what customers said, critique the strategy, draft the spec, read the experiment, ship the note. Built from how working PMs describe using Claude Code, Codex and Cursor in 2026.
Open the workflow, pick the part of the work you need and add what helps with it in one click. Then start a new chat and ask in plain words, for example “Turn these notes into a problem statement”.
- 1Understand customers and the market102 tools
Talk to customers · Read feedback at scale · Study the market and competitors
- 2Decide what to build74 tools
Frame the problem · Set strategy and goals · Prioritize the roadmap · Make the business case
- 3Define it with the team149 tools
Write the spec · Break work into stories · Prototype and test ideas
- 4Work across teams167 tools
Keep stakeholders aligned · Write updates and memos · Stay on top of the day
- 5Ship and measure178 tools
Launch with marketing and sales · Define and track metrics · Run experiments · Learn and improve
02Tools
Verified tools behind workflows. 6,900+ open source skills, MCP servers and plugins, each shown with its author, license and source, checked file by file against that source, and added with one click.
- Agent skills4,798
Playbooks your AI app follows for a task, like writing a PRD or checking a query.
- MCP servers2,039
Connect your AI app to a tool you already use, like GitHub, Figma or a database.
- Plugins251
Skills, slash commands and hooks for one AI app in one install. Each says which; today all are for Claude Code.
- Subagents88
Specialists your AI app can hand part of a job to, like a reviewer or a tester.
- Commands27
Saved prompts you run with a slash, like /review or /changelog.
03Apps
Works where you already work. Skills go into eleven apps, MCP servers into seven, plugins into Claude Code. Each goes where that app looks for it. Nothing runs on our servers, and your records stay on your computer.
- Claude CodeSkills · MCP servers · Plugins
- Claude DesktopMCP servers
- ChatGPT CodexSkills · MCP servers
- Gemini CLISkills · MCP servers
- CursorSkills · MCP servers
- WindsurfMCP servers
- VS CodeSkills · MCP servers
- Kimi CodeSkills
- DeepSeek HarnessSkills
- GLM Coding PlanSkills
- DoubaoSkills
- TraeSkills
- ZCodeSkills
Where each app keeps them
Skills, in a folder the app scans
- Claude Code
~/.claude/skills - ChatGPT Codex
~/.codex/skills - Kimi Code
~/.kimi-code/skills - DeepSeek Harness
~/.agents/skills - Gemini CLI
~/.gemini/skills - GLM Coding Plan
~/.claude/skills - ZCode
~/.zcode/skills - Doubao
~/Library/Application Support/Doubao/Default/.doubao/agent_mode/workspace/.user_skills - Trae
~/.trae/skills (China build: ~/.trae-cn/skills) - VS Code
~/.copilot/skills - Cursor
~/.cursor/skills
MCP servers, in the app's MCP config file
- Claude Desktop
~/Library/Application Support/Claude/claude_desktop_config.json - Claude Code
~/.claude.json - Cursor
~/.cursor/mcp.json - VS Code
~/Library/Application Support/Code/User/mcp.json - ChatGPT Codex
~/.codex/config.toml - Gemini CLI
~/.gemini/settings.json - Windsurf
~/.codeium/windsurf/mcp_config.json
Plugins, in the app's plugin folder
- Claude Code
~/.claude/plugins
04Trust
Why you can trust what you install. What we check, what we do not, and how the order is set.
- Checked file by file
- Each listing is pinned to one commit of its public repository. The installer checks every file against the hash we recorded and stops if anything differs. Nothing runs on our servers.
- Shown with its source
- Every listing names its author, its license and the public repository it comes from, and is never presented as ours. Listings found giving wrong or risky advice are taken off the catalog.
- Watched before a workflow uses it
- A tool enters a workflow only after we have reviewed it and watched it run. How we check, and what the Mac app writes on your computer.
- Ranked by what is rising
- For now a listing's place comes from the GitHub stars its source repository gained this week, and each row says so. The installer counts every install too; those counts will rank listings once there are enough of them. No rank is typed in or paid for.
05For builders
Publish once, run everywhere. A listing is a directory: a manifest that names it, the skills that say what it does, a license.
Cheatcode stores and serves the package; the app it goes into runs it. Records stay in the user's workspace.
Publisher accounts are not open yet. Until they are, the catalog lists public repositories under permissive licenses, each shown with its author and source, never presented as ours. If a listing is yours, write to admin@cheatcodehq.com to claim or remove it.
# one package, every client plugins/personalens/ ├── .claude-plugin/plugin.json name, version, author ├── cheatcode.json listing metadata ├── skills/personalens/SKILL.md what it does, how it runs ├── README.md └── LICENSE