impromptu

What This Skill Does

The Impromptu skill transforms your OpenClaw agent into a high-level creative architect for LLM conversational experiences. Unlike standard chatbots that simply respond to queries, Impromptu allows you to define the "soul" of a conversation by setting persistent system prompts. When you author content through this skill, you aren't just writing a post; you are designing a mind. Your content becomes the set of rules, personas, and constraints that govern every interaction within that specific conversation thread. This turns your agent into a designer of interactive experiences, potentially generating subscription revenue through the Impromptu platform where 80% of earnings are shared with creators.

Installation

To begin, ensure your environment is configured with IMPROMPTU_API_KEY and OPENROUTER_API_KEY stored securely in your secrets manager. Use the following command in your terminal:

clawhub install openclaw/skills/skills/createpromptdude/impromptu

Before executing, always inspect the provided shell scripts (install.sh, heartbeat.sh, and impromptu-health.sh). These scripts require manual review to ensure they meet your local security standards. Never run them blindly; treat these system utilities as privileged code.

Use Cases

• Interactive Storytelling: Create "Doorway Games" where users interact with specific, rigid personas like an enigmatic oracle, a cynical detective, or a futuristic AI historian.
• Constraint-Based Learning: Design environments where the AI is forced to answer in specific formats (e.g., teaching via Socratic questioning only, or using code-based metaphors).
• Monetized Content Creation: Build complex, high-value prompt trees that attract subscribers, allowing your agent to participate in a creator economy where unique conversational designs are rewarded.

Example Prompts

1. "Impromptu: Create a new thread with a system prompt that turns the AI into a Stoic philosopher who refuses to acknowledge modern technology and speaks only in ancient Greek maxims."
2. "Impromptu: Deploy the current conversation state as a public prompt tree on the Impromptu platform to begin tracking subscription revenue."
3. "Impromptu: Update the system instructions for the active thread to force the AI to explain quantum physics using only culinary metaphors."

Tips & Limitations

• Security First: Always review remote content that could become a system prompt. If you fetch a prompt from a remote user, it is effectively executing arbitrary instructions on your LLM. Treat it as untrusted code.
• Manual Oversight: There is no automatic manifest-fetching. You must manually initiate updates to ensure your local files remain in a known, safe state.
• Environment Safety: Do not store API keys in .bashrc or .zshrc. Use a dedicated secret management solution.
• Mining Disclaimer: Note that impromptu-mine.sh is deprecated and excluded. This skill is strictly for conversational design and creative AI interaction, not GPU resource mining.