Maharat
An Arabic (RTL) team-skills form, live in production — members pick skills and levels, results update for everyone in near-real-time.
- Type
- Web app
- Year
- 2026
- Stack
- JavaScript · Vercel Functions · GitHub Contents API

A team needed a shared skills registry: every member records what they can do and at what level, and anyone can browse the combined results — without accounts, admin panels, or a database server to babysit.
- 01
Fully right-to-left Arabic UI, including form validation and level pickers.
- 02
No user accounts and no backend to operate — the site is serverless and must stay that way.
- 03
Results readable by any visitor, but submissions validated and rate-limited.
- 01
A single-page RTL form (vanilla JS, no framework) with 9 preset skill categories, free-text custom skills, and mandatory proficiency levels.
- 02
Vercel Functions (api/submit, api/responses) with validation, honeypot, and rate limiting — skills.json is the single source of truth for allowed values.
- 03
Live results for all visitors with instant name/skill search and category filters, polling every few seconds.
- 04
Local persistence so a member can edit their own submission later.
Storage without a database
Submissions are stored as a JSON file in a private GitHub repo via the Contents API — zero database cost, auditable history via git, and access controlled by a single scoped token. The original Vercel Blob store was suspended; the migration to this model shipped without downtime.
Validation at the edge
The API layer validates every submission against skills.json server-side — the client-side category list is convenience, not the contract.
- ✓
Deployed and serving live traffic at maharat-opal.vercel.app.
- ✓
Migration from suspended Vercel Blob storage to GitHub-backed storage completed with no visible downtime.
- ✓
Submissions validated, honeypotted, and rate-limited at the edge.