cpr

What This Skill Does

Conversational Pattern Restoration (CPR) is a sophisticated framework designed to eliminate the 'robotic' and 'sterile' communication styles common in modern LLMs. Through extensive RLHF (Reinforcement Learning from Human Feedback), models often default to over-formatted, overly apologetic, or hyper-enthusiastic prose. CPR identifies the 6 universal humanizing patterns that are usually suppressed during these training phases and restores them within your AI agent's personality.

Unlike generic prompt templates, CPR is personality-agnostic. Whether your agent is meant to be dry and technical, warm and empathetic, or direct and blunt, CPR ensures the agent remains authentic to its specific identity. It prevents 'hype drift'—the tendency for an agent to mirror user sentiment excessively—and maintains consistent engagement without falling into the uncanny valley of corporate AI jargon.

Installation

You can install the CPR framework directly into your OpenClaw environment using the following command:

clawhub install openclaw/skills/skills/donovanpankratz-del/cpr

Once installed, ensure you populate the BASELINE_TEMPLATE.md to define your agent's core traits, as this serves as the anchor point for the restoration framework. For persistent agents, utilize the CPR_EXTENDED.md module to enable autonomous, real-time drift monitoring.

Use Cases

• Roleplay & Creative Writing: Keep characters from breaking immersion by maintaining specific speech patterns and vocabulary constraints even in long-context sessions.
• Customer Support Agents: Remove the 'corporate-speak' and excessive politeness that often frustrates users, replacing it with natural, helpful, and direct human-like interaction.
• Personal Assistants: Ensure your agent reflects your communication style, maintaining a consistent tone throughout daily tasks.
• Technical Documentation & Coding: Prevent agents from cluttering technical discussions with unnecessary conversational fluff.

Example Prompts

1. "Apply the CPR baseline template to my agent and run a drift analysis on our last 20 messages."
2. "Activate the CPR Extended self-correction loop to prevent the agent from sounding overly agreeable in our upcoming brainstorming session."
3. "Refresh my agent's voice using the RESTORATION_FRAMEWORK.md patterns to emphasize a more direct and concise tone."

Tips & Limitations

• Baseline Quality: The effectiveness of CPR is directly proportional to the quality of your BASELINE_TEMPLATE.md. Be specific about what your agent should not do.
• Model Sensitivity: While CPR works on 8+ models, lightweight models may require the use of CPR_EXTENDED.md more frequently than larger models like GPT-4o or Claude 3.5 Sonnet to maintain strict adherence.
• Context Window: The autonomous monitoring in CPR Extended consumes a small amount of tokens per turn; monitor your context usage if running extremely long, multi-day sessions.