AI Prompt Engineering, Automated

Stop writing prompts.
Start gilding them.

Gild turns rough instructions into structured, scored, deployment-ready prompts using a self-critiquing AI pipeline. Built for people who use AI for real work.

Get Gild for Windows See how it works →
4-stage AI pipeline
Scored & deployment-ready output
Your API key, sent direct to Anthropic

What Gild does to your prompt

Your rough idea
Python security vulnerability reviewer
91
score
Gild output
You are a senior Python security engineer specializing in OWASP Top 10 vulnerabilities. Analyze the provided function and return a structured report: (1) Security findings — CWE ID, exploit scenario, corrected snippet; (2) Performance issues — complexity estimate and optimized version; (3) PEP8 violations with fixes. DO NOT invent CVE IDs or line numbers you cannot confirm.
The Pipeline

Four stages.
One great prompt.

Stage A

Intake &
Clarify

Gild reads your raw idea and extracts the goal, audience, constraints, and output format even when you haven't spelled them out. No intake form. No back-and-forth.

Stage B

Generate

Claude Sonnet drafts a structured, precise prompt with clear role definitions, behavioral constraints, output specs, and edge-case handling baked in by default.

Stage C

Refine

The pipeline turns on itself: a second AI pass audits the draft for ambiguity, gaps, and failure modes, then rewrites until the issues are resolved, not just noted.

Stage D

Final

Every output gets a score across 7 dimensions: clarity, context, specificity, structure, robustness, actionability, and domain optimization. Plus deployment notes on where to paste it and what temperature to run it at.

What You Get

Everything a prompt
should be.

Scored output, every time

Every prompt leaves Gild with a score across 7 named dimensions: clarity, context, specificity, structure, robustness, actionability, and domain optimization. You know exactly why it scored what it did, and what to fix.

Deployment notes included

Gild tells you exactly where to use the prompt, how to tune it for different models, and what edge cases to watch for. No guesswork on your end.

Seconds, not sessions

What takes most people an hour of iteration (drafting, testing, fixing) Gild compresses into a single pipeline run. Write once, deploy with confidence.

Self-critiquing pipeline

Stage C doesn't just review. It rewrites. The pipeline catches its own ambiguities and resolves them before the output ever reaches you.

Edit and re-score

Tweak the refined prompt yourself and re-run Stage D to get a fresh scorecard. See exactly how your edits move the needle, dimension by dimension.

Your runs, saved locally

Run history stays on your device. Gild has no backend — requests go direct from the app to Anthropic over HTTPS using your own API key. Runs typically cost $0.05–$0.30 in your own credits.

Early Access

Built for people who
ship with AI.

You're an early adopter

Gild is new, but the value is easy to verify: run it on one real prompt and compare the before and after. The pipeline's quality is the proof.

30-day money-back guarantee

Run the pipeline on your real use cases. If it doesn't deliver better prompts, scored and deployment-ready, get a full refund. No questions.

Feedback shapes the roadmap

Early users get direct access. If something's missing or broken, say so. The pipeline, the scoring, and the domain coverage are all actively improving.

Live Example

Built with Gild.
The proof is this page.

One rough sentence went into Gild. It came out as a scored, deployment-ready audit prompt — 95/100. We ran it through Claude Opus 4.8, and everything it found is now live on this page.

The prompt Gild generated
Input i need an extensive professional graphic/website designer audit on my digital product landing page.
Domain UI/UX · Graphic Design
Score 95/100
Grade A
Generated 6/30/2026, 4:42 PM
Scorecard
Clarity
19/20
Context
14/15
Specificity
14/15
Structure
15/15
Robustness
14/15
Actionability
10/10
Domain Fit
9/10
Deployment note

Best deployed as a system prompt in a multimodal-capable assistant platform that accepts both file uploads and image attachments. Temperature 0.2–0.4. Web search disabled to enforce anti-fabrication discipline.

Here's the loop. We gave Gild one rough sentence: "I need a professional design audit of my landing page." Gild turned it into the prompt scored above. We then ran that prompt through Claude Opus 4.8 — and the fixes it surfaced (text contrast, heading structure, CTA labels, the trimmed pricing card) are what you're looking at now. Gild wrote the prompt. Opus 4.8 ran it. The page is the receipt.
Pricing

One plan.
Everything included.

Annual License
Gild Pro
$99
per year  ·  ≈ $8/mo  ·  cancel anytime
  • Unlimited pipeline runs
  • Scored, deployment-ready output every time
  • Self-critique & rewrite pass on every prompt
  • Native Windows app — fast, no browser tab
  • All future updates included for the year
Get Gild — $99/yr
Secure checkout via Lemon Squeezy
Bring your own Anthropic API key — runs cost ~$0.05–$0.30 each in your own credits. Single-user license, 2 device activations. Windows installer + key emailed on purchase. 30-day money-back guarantee.
Requires: Windows 10/11  ·  Anthropic API key  ·  Internet connection for pipeline runs  ·  macOS coming soon
FAQ

Questions,
answered.

Do I need an Anthropic API key?
Yes. Gild runs on your own Anthropic API key. You'll need an active key from console.anthropic.com to use the app.
Can I use my Claude Pro/Max subscription instead?
No. Claude Pro and Max subscriptions are separate from API access. Gild requires a standard Anthropic API key, billed per use through the Anthropic Console.
Where is my API key stored?
In your OS credential vault (Windows Credential Manager). It never touches Gild's servers (we don't have any) and is never transmitted anywhere except directly to Anthropic.
Does Gild store my prompts?
No. Your prompts and pipeline runs stay local on your machine. Nothing is sent to us or stored on any server we control.
What happens after I buy?
You'll receive the Windows installer and a license key by email immediately after purchase, ready to download and activate.
Can I install on more than one machine?
Yes. Each license includes up to 2 device activations, so you can use Gild on both your desktop and laptop.
Does it work on macOS?
Not yet. Gild is currently Windows only. macOS support is planned.
What does the 30-day guarantee cover?
A full refund if you're not satisfied, no questions asked. Just reach out within 30 days of purchase.

Your prompts deserve
to be gilded.

Stop guessing. Start shipping prompts that actually work.

Get Gild