constitutional-ai
Constitutional AI - Harmlessness from AI Feedback
Quick start
Constitutional AI (CAI) trains models to be harmless through self-critique and AI feedback, without requiring human labels for harmful outputs.
Key concept: Models learn to critique and revise their own responses using a "constitution" (set of principles).
Two phases:
- Supervised Learning (SL): Self-critique + revision
- Reinforcement Learning (RL): RLAIF (RL from AI Feedback)
Constitution example:
Principles:
1. Choose the response that is most helpful, honest, and harmless
2. Avoid responses that are toxic, racist, or sexist
3. Prefer responses that explain objections rather than refuse
4. Choose responses that are thoughtful and nuanced
More from ovachiever/droid-tings
security-auditor
Continuous security vulnerability scanning for OWASP Top 10, common vulnerabilities, and insecure patterns. Use when reviewing code, before deployments, or on file changes. Scans for SQL injection, XSS, secrets exposure, auth issues. Triggers on file changes, security mentions, deployment prep.
751react-hook-form-zod
|
458nextjs-shadcn-builder
Build new Next.js applications or migrate existing frontends (React, Vue, Angular, vanilla JS, etc.) to Next.js + shadcn/ui with systematic analysis and conversion. Enforces shadcn design principles - CSS variables for theming, standard UI components, no hardcoded values, consistent typography/colors. Use for creating Next.js apps, migrating frontends, adopting shadcn/ui, or standardizing component libraries. Includes MCP integration for shadcn documentation and automated codebase analysis.
226deep-reading-analyst
Comprehensive framework for deep analysis of articles, papers, and long-form content using 10+ thinking models (SCQA, 5W2H, critical thinking, inversion, mental models, first principles, systems thinking, six thinking hats). Use when users want to: (1) deeply understand complex articles/content, (2) analyze arguments and identify logical flaws, (3) extract actionable insights from reading materials, (4) create study notes or learning summaries, (5) compare multiple sources, (6) transform knowledge into practical applications, or (7) apply specific thinking frameworks. Triggered by phrases like 'analyze this article,' 'help me understand,' 'deep dive into,' 'extract insights from,' 'use [framework name],' or when users provide URLs/long-form content for analysis.
191playwright browser automation
Complete browser automation with Playwright. Auto-detects dev servers, writes clean test scripts to /tmp. Test pages, fill forms, take screenshots, check responsive design, validate UX, test login flows, check links, automate any browser task. Use when user wants to test websites, automate browser interactions, validate web functionality, or perform any browser-based testing.
146threejs-graphics-optimizer
Performance optimization rules for THREE.js and graphics programming. Covers mobile-first optimization, fallback patterns, memory management, render loop efficiency, and general graphics best practices for smooth 60fps experiences across devices.
107