What does this skill do?

The Skill Recommendation Pipeline Architect designs complete systems that select the best K items for a user or context, using a six-stage pattern: Source, Hydrator, Filter, Scorer, Selector, and SideEffect. It generates architecture, explicit design trade-offs, and executable code ready to run in Python, Go, or TypeScript.

Social Media Feed
It combines on-network and off-network candidates, scores the likelihood of a like, comment, or skip, and serves the top K with asynchronous side effects.
Task Prioritizer
Filter completed or expired tasks, rank them by urgency and effort, and return the 10 highest-priority tasks with asynchronous logging.
RAG Reranker
Reorder documents retrieved by an RAG system by applying eligibility filters and multi-action scoring with adjustable weights.
Notification Triage
Select the most relevant notifications for a user by applying filters based on age and previous views, using a combined scoring system.

Usage examples

📱 Social Feed
Design a social feed pipeline in TypeScript that combines posts from followers and trending topics, scores likes/comments/skips, and returns the top 50.
📋 Task Prioritizer
Create a pipeline in Python using FastAPI that prioritizes tasks by urgency and effort, filtering out completed and expired tasks.
🔍 RAG Reranker
Design a reranker for my RAG system that filters out duplicate documents and reorders them based on relevance combined with diversity.
⚙️ Multi-action with weights
Migrate my single-relevance scorer to a multi-action system with runtime-adjustable weights in Go.

Features

Six-stage framework Source→Hydrator→Filter→Scorer→Selector→SideEffect: a composable pattern for any item selection system.
Work Scaffolding Generate code that's ready to run in Python, Go, or TypeScript—not pseudocode or abstract diagrams.
Multi-Action Scoring Combine multiple predictions (like, comment, skip) with weights that can be adjusted at runtime.
Asynchronous side effects The cache of served IDs, print events, and analytics runs on a "fire-and-forget" basis without blocking the response.
Explicit trade-offs Document design decisions such as standalone scoring vs. combined scoring, or single score vs. multi-action.

Frequently asked questions

Generate executable scaffolding in TypeScript, Go, and Python. Specify your tech stack when describing the use case.
Not necessarily. The skill designs the infrastructure around the scorer; you can use an existing model or define heuristic rules.
No. All side effects (cache, events, analytics) are implemented asynchronously using a "fire-and-forget" approach.
Yes. The six-step framework can be applied to any system that needs to select the best K items: feeds, notifications, RAG rerankers, and task prioritizers.

Related skills

More skills in the same category:

Recommendation Pipeline Architect — Designing Ranking Systems with Claude AI

¿Prefieres escuchar el contenido? Genera la narración de audio con un clic.