Claude Prompts MCP
MCP server for prompt templates, step chains and output validation, exported as skills for Claude Code, Cursor and Codex
- 192 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Gates with a shell command run code on your machine
- The agent creates and edits templates that later steer its own work
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/claude-promptsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Gates with a shell command run code on your machine
- The agent creates and edits templates that later steer its own work
This is third-party code. Review the repository files before installing.
What it does
The server adds a reusable workflow layer to the client. There are four elements. YAML prompt templates with named arguments are called with `>>review target:'src/auth/'`. Gates (the `::` mark) are checks the agent applies to its own answer, blocking or advisory, and they can run a shell command such as `:: verify:"npm test"`. Reasoning frameworks (the `^` mark) like `^ReACT` shape the work and styles (the `#` mark) set format and tone. The bundle holds 52 prompts in nine categories and the agent writes new ones itself through the `resource_manager` tool. The three MCP tools are `prompt_engine`, `resource_manager` and `system_control`. Templates hot-reload on save and have versions and rollback.
Who it is for. Developers and teams who repeat the same prompts, build multi-step workflows and want to share them across Claude Code, Cursor and Codex.
Good fit when
- The same code review prompt is retyped and you want to call it with `>>review`
- A study runs in several steps and context must pass from step to step in a chain
- An agent answer is checked with a `::` criterion or `npm test` and a failed step is retried
- A workflow is written once and exported as a skill for Claude Code and Cursor with `skills:export`
- A team prompt library needs versions, history and rollback
Not a fit when
- The built-in `/` commands of your client cover daily tasks
- You want a ready prompt collection without authoring workflows: an ordinary library fits
- The client has no hooks: the `>>`, `::` and `^` marks do not work in conversation there and only the three MCP tools remain
Example request
>>tech_evaluation_chain library:'zod' context:'API validation'MCP
- Transport
- stdio
- Authentication
- not required
Environment variables
MCP_RESOURCES_PATH- Folder with your prompts, gates and frameworks.
Limitations
The `>>`, `::` and `^` marks work in conversation only where hooks are connected: fully in Claude Code and OpenCode, experimentally in Codex and partly in Gemini CLI. In Claude Desktop, Cursor, VS Code and Windsurf only the three MCP tools remain. Codex needs CLI 0.117 or newer and is unavailable on Windows, and hooks are enabled in `config.toml` and trusted with `/hooks`. Gates with a shell command run code on your machine. Styles are managed with the separate `cpm` CLI.
How to disable. In Claude Code run `/plugin uninstall claude-prompts@minipuft`; in other clients remove the server entry from the MCP configuration.
FAQ
How do I install Claude Prompts in Claude Code?
Run `/plugin marketplace add minipuft/minipuft-plugins` once, then `/plugin install claude-prompts@minipuft` and try `>>tech_evaluation_chain library:'zod' context:'API validation'`.
Which clients are supported?
Eight: Claude Code, OpenCode, Codex, Gemini CLI, Claude Desktop, Cursor, VS Code and Windsurf. The full set of marks works with hooks and the rest keep the three MCP tools.
What do the `>>`, `::`, `^` and `#` marks do?
`>>` calls a template, `::` sets an answer check, `^` attaches a reasoning framework and `#` picks an output style.
How do I create my own prompt?
Run `--init=~/my-prompts` to scaffold a workspace with three examples and `config.jsonc`. Edit prompts in YAML or have the agent author them through `resource_manager`.
How do I export a workflow as a skill for another client?
With the `skills:export` command from the package: one YAML workflow becomes the native skill format of the chosen client.
Discussion
No questions or reviews yet.
Related
MCP serversA prompt template store with search and variable filling, plus a script that runs Claude agents across a whole project.
Open prompt library with a Claude Code plugin, MCP server and CLI: search, fetch and improve prompts and skills from an agent
Claude Code Prompts (RepoWise)
claude-code-prompts
Independently written prompt templates for AI coding agents: system prompts, tool prompts, delegation, memory and multi-agent coordination
PromptOps keeps Claude Code, Codex and Gemini CLI side by side by project and gives a lead agent MCP to launch helpers
The author's personal Claude Code toolbox: meta-prompt and todo commands, skills to create skills, hooks and MCP servers, auditors and a Ralph loop
An MCP server and editor: a prompt is split into blocks, edited as a flowchart and compiled into XML for Claude
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Claude Prompts (this card) | no | not required | 192 | |
| mcp-prompts | no | not required | 120 | stores and serves prompts; this one adds step chains, output gates and skill export |
| just-prompt | no | key is optional | 736 | sends one prompt to several models; this one has templates, chains and gates inside the client |
README in short
The README shows a quick start for Claude Code and Codex, expandable setup for other clients and a table of four elements: prompts, gates, frameworks and styles. Then come chain composition examples, run modes and documentation links.