ericcgladstone-maker

What Should I Ride — Data skill for Claude Code

Data community

Bicycle tire recommendation engine built on a two-stage Claude reasoning pipeline.

How to install What Should I Ride

This entry records only its repository, not the path inside it, so there is no exact command to give. Open ericcgladstone-maker/what-should-i-ride and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What What Should I Ride does

Bicycle tire recommendation engine built on a two-stage Claude reasoning pipeline. Returns ranked recommendations with per-tire confidence scoring, a paired front and rear suggestion, and a mechanic memo.

Alternatives in Data

Notion Sync — /notion-sync - Push Ranked Jobs and Applications to a Notion Database 36.6k ★
Claude Deep Research Skill — Enterprise-grade deep research skill for Claude Code with 8-phase pipeline, source credibility scoring, and au 1k ★
Graph Engineering — Graph engineering for AI agents: the 9-stage knowledge-graph pipeline (translated from SEU's graduate course) 452 ★

README

What Should I Ride?

An AI-powered bicycle tire recommendation engine built on a two-stage Claude reasoning pipeline. Riders describe their situation in natural language; the system returns structured, ranked recommendations with per-tire confidence scoring, a paired front/rear suggestion, and a plain-text mechanic memo.

Live demo: deployed as a Netlify static site — no server, no database, no build step.

Architecture Overview

The application is a single React 18 component rendered from a modular file structure. No JSX, no bundler, no transpilation. It deploys by dropping a folder.

index.html          Entry point — loads dependencies and modules in order
css/main.css        All styles. CSS variables in :root for easy theming.
js/config.js        Constants: API config, image paths, option lists, PRODUCT_CATEGORY
js/prompts.js       The four AI system prompts — auditable and editable independently
js/archetypes.js    12 buyer archetypes, form configs, carousel prompt library
js/helpers.js       Stateless utilities and sub-components (no side effects)
js/app.js           React App component — all state and rendering
vendor/             React 18.2.0 and ReactDOM 18.2.0 UMD production builds (vendored locally)
images/             16 static assets (hero images + archetype tiles)

Load order matters: each file depends on variables declared before it. `app.js` is last and calls `ReactDOM.createRoot` to mount.

React and ReactDOM are vendored locally in `vendor/` — no CDN dependency at runtime.

The AI Pipeline

Every recommendation runs two sequential API calls to Claude.

Stage 1 — Diagnostic Pre-Pass (`DIAG_PROMPT`)

Receives raw rider input. Returns structured JSON:

Field	Purpose
`realProblem`	The actual mechanical/performance problem, or decision type
`characteristicVector`	3-5 tire characteristics needed, in priority order
`surfacePrior`	What named location/trail/race implies about terrain and local tir