✨ Prompt Polisher¶

Live at: aguidetocloud.com/prompt-polisher/ Built: April 2026 Cost: $0 — 100% client-side JavaScript, zero API calls

What It Does¶

An interactive prompt improvement tool that scores any AI prompt on a 0–100 scale using the CRAFTS framework, then automatically rewrites it with missing elements — all running entirely in the browser with no backend.

Think of it like Grammarly, but for AI prompts — paste a weak prompt, get a stronger one instantly.

Architecture Overview¶

┌──────────────────────────────────────────────────────────┐
│              100% Client-Side (No API Calls)              │
├──────────────────────────────────────────────────────────┤
│                                                          │
│  User Input                                              │
│     ↓                                                    │
│  Domain Detection (tech/business/creative/academic/      │
│                    education/email/data)                  │
│     ↓                                                    │
│  CRAFTS Scoring (6 pillars, 100 points total)            │
│     ↓                                                    │
│  Smart Rewrite Engine (adds missing elements)            │
│     ↓                                                    │
│  Output: Polished prompt + score delta + diff + tips     │
│                                                          │
├──────────────────────────────────────────────────────────┤
│  layouts/prompt-polisher/list.html  (Hugo template)      │
│  static/js/polisher.js             (936 lines)           │
│  static/css/polisher.css           (coral accent theme)  │
└──────────────────────────────────────────────────────────┘

The CRAFTS Scoring Framework¶

Each prompt is scored across 6 pillars totalling 100 points:

Pillar	Points	What It Checks
Context	15	Background info, situation, constraints
Role	20	Who the AI should act as (persona/expertise)
Action	25	Clear task/verb — what to do (highest weight)
Format	15	Output structure (list, table, email, code, etc.)
Tone	10	Communication style (professional, casual, etc.)
Scope	15	Boundaries, length, audience, exclusions

Score Labels¶

Range	Label
0–20	Needs Work
21–40	Getting There
41–60	Good Start
61–80	Strong
81–100	Expert Level

Scoring is heuristic/keyword-based — it checks for patterns, keywords, and structural elements in the prompt text. No AI model is involved in scoring.

Smart Rewrite Engine¶

When a pillar scores low, the rewrite engine adds the missing element using domain-aware suggestions:

Missing Pillar	What Gets Added
No Role	Adds "Act as a [domain-specific expert]..."
No Context	Adds background/scenario relevant to the detected domain
No Format	Adds output structure (e.g., "Format as a numbered list...")
No Tone	Adds tone guidance (e.g., "Use a professional tone...")
No Scope	Adds boundaries (e.g., "Keep it under 500 words...")

7 Supported Domains¶

The engine detects the domain from keywords in the prompt:

tech · business · creative · academic · education · email · data

Each domain has tailored suggestions (e.g., tech prompts get "Act as a senior software engineer", business prompts get "Act as a management consultant").

Additional Rewrites¶

Strips vague openers ("Can you help me...")
Capitalises first letter
Adds punctuation if missing
Expands very short prompts
Appends platform-specific hints when a platform is selected

Frontend Features¶

Feature	Detail
Score dashboard	Animated count-up + per-pillar progress bars
Green diff highlighting	Added text shown in green (`.polisher-added` class)
Before/After toggle	Side-by-side comparison view
Re-score delta	Shows "Original: 22 → Polished: 78 (+56 points)"
Platform-aware rewrites	Dropdown: Any AI / M365 Copilot / ChatGPT / Claude / Gemini
5 example prompts	Pre-loaded examples to try the tool
Related prompts	3 prompts from the Prompt Library matched by domain
localStorage history	Last 10 polished prompts saved locally
Ctrl+Enter shortcut	Quick-polish keyboard shortcut (hidden on mobile)
"Edit & re-polish"	Scrolls back to input for iterating
3-step "How it works"	Pre-analysis guide, hidden after first use
CRAFTS explainer	Full breakdown of each pillar with examples

Accessibility¶

Feature	Implementation
`aria-live="polite"`	Results region announces updates to screen readers
`aria-label`	Input textarea + platform select labelled
`tabindex="0"` + `role="button"`	History items keyboard-navigable
`Enter`/`Space` support	History items respond to keyboard
`focus-visible` styles	Visible focus outlines on all interactive elements
`prefers-reduced-motion`	Disables animations for motion-sensitive users

SEO¶

JSON-LD WebApplication schema in the Hugo template
Body class: page-polisher for page-specific CSS

CSS — `polisher.css` (Coral Accent `#FF6B6B`)¶

Element	Detail
Accent	Coral `#FF6B6B`
Score bars	Animated width transitions per pillar
Green diff	`.polisher-added` = green text + background
Compare view	Before (red tint) / After (green tint) panels
History	Clickable items with hover/focus states
Mobile	Ctrl+Enter hint hidden, stacked layout

Key Design Decisions¶

Decision	Rationale
Zero API calls	Runs entirely in browser — no cost, no latency, no privacy concerns
Heuristic scoring	Fast, predictable, and transparent — users understand why they scored low
Domain detection	Tailored rewrites feel relevant, not generic
CRAFTS acronym	Memorable framework that teaches prompt engineering principles
Cross-links to Prompt Library	"Here are 3 prompts related to your domain" — drives discovery
localStorage history	Users can revisit and re-polish previous prompts

File Inventory¶

File	Lines	Purpose
`layouts/prompt-polisher/list.html`	~190	Hugo template — UI structure + JSON-LD
`static/js/polisher.js`	936	Core logic — scoring, rewriting, UI
`static/css/polisher.css`	~670	Styling — coral accent theme
`content/prompt-polisher/_index.md`	5	Front matter only (title/description/layout)

Remaining TODOs¶

Feature	Status
Teach Me interactive walkthrough	🔲 Planned
Share Score card (social sharing)	🔲 Planned
Custom OG image for social previews	🔲 Planned

Last updated: 11 April 2026