school-analytics

name	school-analytics
description	Fetch and analyze OpenCode School enrollment and progress data from Cloudflare KV

When to use

Use this skill when asked about enrollments, student counts, progress stats, lesson completion rates, exercise completions, demographics, or any analytics about OpenCode School students.

Requirements

Requires the Cloudflare MCP server (github.com/cloudflare/mcp). The cloudflare-api_execute tool provides a pre-set accountId variable and a cloudflare.request() helper for authenticated API calls.

Data source

Student records are stored in Cloudflare KV. To find the namespace ID, read wrangler.jsonc from the project root and find the KV namespace with binding PROGRESS. Use its id value.

Fetching all keys

Use the Cloudflare API, not the wrangler CLI. wrangler kv key list returns only 25 keys and has no pagination flag.

List keys via GET /accounts/{accountId}/storage/kv/namespaces/{namespaceId}/keys with a limit of 1000. The response includes result_info.cursor for pagination -- keep fetching until the cursor is absent or the result set is empty. Each key object has a name field.

Fetching all values

Fetch each value via GET /accounts/{accountId}/storage/kv/namespaces/{namespaceId}/values/{key} (URL-encode the key). The API returns the value as a JSON string, not a parsed object -- always JSON.parse() the result. Fetch in parallel batches of ~20 to stay under rate limits.

Record schema

Each student record has these fields:

• createdAt -- ISO 8601 string, set at enrollment time
• updatedAt -- ISO 8601 string, updated on every write
• completedLessons -- array; items are either bare strings (legacy format) or objects with slug, completedAt, source, and optional model. Normalize by extracting the slug field from objects.
• completedExercises -- same shape as completedLessons
• profile -- optional object with fields: codingExperience, aiTools, editor, terminalComfort, learningStyle, depthPreference, languages, os
• deviceId -- optional UUID string, present only if the client sent one at enrollment

Aggregate inside the sandbox

The cloudflare-api_execute tool truncates responses over ~6000 tokens. Always compute aggregates (counts, buckets, funnels) inside the sandbox function. Never return raw student records.

Common analyses

• Time-window filtering: compare createdAt against a cutoff timestamp
• Hourly/daily enrollment rates: bucket by the date or date-hour portion of createdAt
• Lesson completion funnel: count how many students completed each lesson slug
• Exercise completions: count how many students completed each exercise slug
• Profile demographics: count by profile.os, profile.editor, profile.codingExperience
• Progress depth: bucket students by number of lessons completed (0, 1-3, 4-7, 8+, all)