Bring your prompt library into your AI assistant.

Start a chat in your local agent and paste this message. It identifies its client, installs the Skill, and configures the connection.

Help me connect PicPromptLab to this AI assistant (Codex, Claude Code, or Cursor):
1. Identify the client. Read and inspect https://stunprompt.com/skills/picpromptlab-create/SKILL.md, then install this Skill in my user-level skills directory.
2. Configure the remote MCP server https://stunprompt.com/mcp for my user account. Preserve my existing MCP connections.
3. Start the client OAuth flow. I will sign in to PicPromptLab and approve access myself in the opened page. Do not ask for my password or tokens, and do not approve access on my behalf.
4. Verify that search_prompts and generate_image are available, then tell me whether a new session is needed. Do not generate images, create prompts, or publish anything during setup.
If you cannot complete a step in this client, give me the exact next action.

Want to review it first? Read the Skill source

After setup, your client opens or points you to the PicPromptLab sign-in and consent page. Review the client name and permissions, then decide whether to allow it. Ask the agent to verify the connection.

Your agent searches for reusable prompts, adapts one in this conversation, and generates a private image. It does not add a prompt entry by default. Saving a reusable draft and publishing each require a separate request.

“Use my prompt library to create a warm product image of a ceramic cup. Show me the result first; do not save a prompt or publish.”

For image edits, your agent can upload a local image as a private asset before generating. It can also manage existing prompts, assets, and collections. The remote service cannot read a local file path directly.