Prismpt

Analyze. Refine. Prompt better.

Prismpt is a modern, open-source developer tool that dissects, evaluates, and optimizes artificial intelligence prompts like a prism splitting white light into its core spectrum.

Rather than relying on vague trial-and-error, Prismpt analyzes your prompt across six fundamental dimensions, detects structural deficiencies, extracts key architectural segments (Role, Task, Context, Constraints, Format), and generates structured, production-ready blueprints—100% client-side with zero telemetry and zero external API dependencies.

---

Features

• 🔬 Local Heuristic Engine: Fast, deterministic, and explainable evaluation across 6 dimensions. Zero external AI API required.
• 🎯 Multi-Dimensional Scoring (0–100): Evaluates Clarity, Context, Specificity, Structure, Constraints, and Intent.
• 🏅 Tiered Grading (S, A, B, C, D, F): Instant visual grading reflecting structural completeness.
• 💡 Actionable Feedback: Concrete Strengths (✓) and Improvements (!) with clear explanations.
• 🏷️ Prompt Structure Highlights: Automatically extracts explicit segments: ROLE, TASK, CONTEXT, CONSTRAINT, and FORMAT.
• ⚡ Prompt Optimizer: Heuristically transforms unstructured text into an optimal blueprint (Role, Objective, Context, Requirements, Output Format, Constraints).
• ⚖️ Before / After Comparison: Side-by-side (desktop) and stacked (mobile) comparison showing prompt improvements and score deltas.
• 🇹🇷 / 🇬🇧 Complete Bilingual Localization: Full Turkish and English interface with auto-detection and language-matching analysis.
• 🔒 100% Privacy-First: Prompts never leave your browser. No remote servers, no analytics, no tracking.
• 📜 Local History: Stores the last 25 prompt analyses in localStorage with reload, delete, and clear-all capabilities.
• 📤 Export & Share: Quick-copy formatted text summaries, or export comprehensive reports as Markdown (.md) or clean JSON (.json).
• 🎨 Developer-First Aesthetic: Inspired by Linear, Vercel, and Raycast with Dark & Light theme support.
• ⌨️ Productivity Shortcuts: Press Cmd/Ctrl + Enter to analyze instantly.

---

Screenshots

Dark Mode (Default)	Light Mode

(To add your own screenshots, place them in the public/ directory).

---

How It Works

Prismpt treats prompt engineering as a disciplined software architecture problem. When you submit a prompt, the engine passes the input through specialized heuristic analyzers:

                  ┌────────────────────────────────────────┐
                  │              Raw Prompt                │
                  └───────────────────┬────────────────────┘
                                      │
               ┌──────────────────────┴──────────────────────┐
               ▼                                             ▼
     [Language Detector]                           [Highlights Extractor]
     (Turkish / English)                       (Role, Task, Context, Format)
               │                                             │
               └──────────────────────┬──────────────────────┘
                                      ▼
               ┌─────────────────────────────────────────────┐
               │         6-Dimensional Analysis Core         │
               ├──────────────────────┬──────────────────────┤
               │ 1. Clarity (18%)     │ 4. Structure (15%)   │
               │ 2. Context (17%)     │ 5. Constraints (16%) │
               │ 3. Specificity (18%) │ 6. Intent (16%)      │
               └──────────────────────┬──────────────────────┘
                                      ▼
               ┌─────────────────────────────────────────────┐
               │    Deterministic Aggregate Score (0-100)    │
               │          Tier Grade (S / A / B / C / D / F) │
               │          Strengths & Improvement Items      │
               └──────────────────────┬──────────────────────┘
                                      ▼
               ┌─────────────────────────────────────────────┐
               │        Prompt Optimizer (Blueprint)         │
               └─────────────────────────────────────────────┘

Every check is explainable and deterministic: identical prompts always produce identical scores.

---

Prompt Scoring

Prompts receive an aggregate score from 0 to 100 and a corresponding letter grade:

Grade	Score Range	Status	Description
S	90 – 100	Excellent	Flawless prompt with expert role, clear task, explicit constraints, and formatting.
A	80 – 89	Great	Highly effective; minor contextual or formatting enhancements possible.
B	70 – 79	Good	Clear directive and baseline structure; would benefit from negative constraints or examples.
C	60 – 69	Needs Work	Understandable but prone to ambiguous hallucinations; missing guidelines.
D	40 – 59	Poor	Weak structure, lacks context, missing output specifications.
F	0 – 39	Inadequate	Very brief, ambiguous, or empty input.

The 6 Evaluation Dimensions

1. Clarity (18%): Readability, imperative directives, absence of rambling filler and ambiguous phrases ("make it cool", "falan filan").
2. Context (17%): Background domain, tech stack, scenario, and input data presence (code blocks, JSON, tables).
3. Specificity (18%): Quantitative metrics, numerical bounds, concrete examples (few-shot), and defined target audience.
4. Structure (15%): Delimiters, markdown headings, bulleted lists, XML tags, and paragraph division.
5. Constraints (16%): Negative boundaries (what not to do), output format specifications (JSON, Markdown table), and tone/length caps.
6. Intent (16%): Expert persona/role assignment, explicit goal formulation, and clear definition of success.

Note

Prismpt's local analyzer uses heuristic rules rather than calling an external LLM. Scores should be regarded as an actionable guidance and structural benchmark rather than an absolute measure of prompt quality.

---

Tech Stack

• Framework: Next.js 15 (App Router, React 19)
• Language: TypeScript (Strict Mode)
• Styling: Tailwind CSS
• Icons: Lucide React
• Animation: Framer Motion
• Testing: Vitest
• Internationalization: Custom zero-dependency React context with localStorage persistence

---

Getting Started

Prerequisites

• Node.js 18.x or later
• npm

Installation

1. Clone the repository:

   git clone https://github.com/bortechin/prismpt.git
   cd prismpt

2. Install dependencies:

   npm install

3. Start the development server:

   npm run dev

4. Open http://localhost:3000 in your browser.

Verification Scripts

# Run unit tests
npm run test

# Run linter
npm run lint

# Build for production
npm run build

---

Project Structure

prismpt/
├── app/
│   ├── globals.css             # Tailwind configuration & CSS variables
│   ├── layout.tsx              # Root layout, metadata & providers
│   └── page.tsx                # Main single-page application orchestrator
├── components/
│   ├── analyzer/
│   │   ├── ExamplePrompts.tsx  # Preset categorized examples
│   │   └── PromptEditor.tsx    # Textarea, counters, shortcuts & toolbar
│   ├── results/
│   │   ├── CategoryBreakdown.tsx # 6 dimension progress cards
│   │   ├── ExportShareModal.tsx  # Text, Markdown, and JSON exporter
│   │   ├── PromptHighlights.tsx  # Detected architectural tags
│   │   ├── ScoreGauge.tsx        # Circular SVG instrument dial
│   │   └── StrengthsIssues.tsx   # Two-column strengths & improvements
│   ├── optimizer/
│   │   └── BeforeAfterView.tsx   # Side-by-side comparison & score delta
│   ├── history/
│   │   └── HistoryModal.tsx      # Local history browser (max 25)
│   ├── about/
│   │   └── AboutModal.tsx        # Methodology & privacy documentation
│   ├── layout/
│   │   ├── Footer.tsx            # Clean developer footer
│   │   └── Header.tsx            # Responsive navigation & switches
│   └── ui/
│       ├── Badge.tsx             # Grade and tag badges
│       ├── Button.tsx            # Button variants with loading states
│       ├── Modal.tsx             # Accessible dialog container
│       └── Toast.tsx             # Toast notification provider
├── lib/
│   ├── analyzer/
│   │   ├── analyzer.ts           # Main scoring orchestrator
│   │   ├── clarity.ts            # Clarity heuristic checks
│   │   ├── constraints.ts        # Constraints heuristic checks
│   │   ├── context.ts            # Context heuristic checks
│   │   ├── highlights.ts         # Structural segment extractor
│   │   ├── intent.ts             # Intent & role heuristic checks
│   │   ├── language.ts           # TR/EN prompt language detector
│   │   ├── specificity.ts       # Specificity & metrics checks
│   │   ├── structure.ts          # Formatting & list checks
│   │   └── types.ts              # Core TypeScript interfaces
│   ├── optimizer/
│   │   └── optimizer.ts          # Structured prompt generator
│   ├── storage/
│   │   └── history.ts            # Client-side localStorage manager
│   ├── i18n/
│   │   └── context.tsx           # Bilingual translation context
│   └── theme/
│       └── ThemeProvider.tsx     # Dark/Light/System theme engine
├── locales/
│   ├── en.ts                     # English translations
│   ├── tr.ts                     # Turkish translations
│   └── types.ts                  # Translation schema
├── public/
│   ├── logo.svg                  # Brand vector logo
│   └── logo-mark.svg             # Favicon & icon mark
├── test/
│   ├── analyzer.test.ts          # Comprehensive analyzer unit tests
│   └── optimizer.test.ts         # Optimizer unit tests
├── .github/
│   ├── ISSUE_TEMPLATE/           # Bug report & feature request templates
│   └── PULL_REQUEST_TEMPLATE.md
├── CHANGELOG.md
├── CONTRIBUTING.md
├── LICENSE                       # MIT License
└── package.json

---

Roadmap

The following enhancements are planned for future releases:

• Optional AI-Powered Deep Analysis: Opt-in provider (OpenAI, Anthropic, Gemini, Ollama) for semantic verification.
• Custom Scoring Profiles: Domain-specific heuristics (e.g., Code Generation, System Prompts, Creative Writing).
• Expanded Language Support: German, Spanish, French, and Japanese localization.
• Browser Extension: Evaluate prompts in ChatGPT, Claude, and Gemini web interfaces.
• Prompt Collections & Tagging: Organize, tag, and search saved prompt templates.
• Shareable Analysis Reports: Exportable standalone HTML reports.
• Prismpt CLI: Run prompt evaluation in CI/CD pipelines to prevent prompt regression.
• VS Code Extension: In-editor prompt analysis for AI engineers.

---

Contributing

Contributions are warmly welcomed! Please read CONTRIBUTING.md for guidelines on branch management, commit styles, and testing requirements.

1. Fork the repo.
2. Create your feature branch (git checkout -b feature/amazing-feature).
3. Verify tests pass (npm run test).
4. Commit your changes (git commit -m 'feat: add support for markdown checklists').
5. Push to the branch (git push origin feature/amazing-feature).
6. Open a Pull Request.

---

Privacy

Prismpt is strictly client-side:

• Your prompts are never transmitted to external servers.
• There are no tracking scripts, cookies, or analytics.
• History is stored exclusively in your browser's localStorage.

---

License

Distributed under the MIT License. See LICENSE for more information.