pref0
por fliellerjulian
Web & Frontend DevelopmentResumen
Learn user preferences from conversations and personalize responses automatically.
Descripción
Qué hace esta skill
pref0 learns user preferences from conversation history and injects them into future sessions automatically. Corrections like "use pnpm, not npm" are extracted, scored by confidence, and served back via a ready-to-use system prompt string. Preferences compound across sessions as the same signals repeat.
Instead of manually maintaining a preferences file or re-stating preferences each session, pref0 extracts and scores them automatically from natural conversation corrections.
Cuándo usarla
- Stop repeating tooling preferences to an AI agent every session
- Capture corrections mid-conversation and apply them going forward
- Inject personalized context into agent system prompts at inference time
- Reset a user's preference profile on request or account deletion
- Build multi-user AI apps where each user gets personalized defaults
Flujo de trabajo de ejemplo
User asks: Help me scaffold a new web project
- 1
Fetch preferences via GET /v1/profiles/{userId}?minConfidence=0.5
- 2
Append the returned prompt field to the system prompt
- 3
Generate scaffold using the learned preferences (TypeScript, pnpm, Tailwind)
- 4
After session ends, call POST /v1/track with the full message history
- 5
pref0 extracts any new corrections or preferences and updates confidence scores
Scaffolded project matches user's known preferences without any explicit instruction; new corrections from the session are stored for next time
Requisitos
- PREF0_API_KEY environment variable
- pref0 account (sign up at pref0.com/signup)
Instalación
clawhub install fliellerjulian/pref0Ejecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.