Deterministic & Auditable

Turn your Brand Manuals into Agent Skills

Transform your brand manuals, style guides, and playbooks into a versioned skill pack – including policies, validators, and tests. Deterministically extracted, with source references for every rule.

What are Agent Skills? Agent Skills are structured rule packages for AI agents: policies, validators, and tests bundled together. Instead of hoping your AI "remembers" brand rules buried in a PDF, skills provide machine-readable policies with built-in validation – so brand compliance becomes a gate, not a hope.

Deterministic extraction, not LLM guessing

Traceable rules with diff tracking

Gates & tests prevent rule drift

Your PDF stays private. Processing runs isolated. Watch every step live.

Why AI agents fail with PDFs

Your brand guidelines are perfectly documented – but AI agents work with context windows. This leads to inconsistencies, rule violations, and endless prompt workarounds.

PDFs are too big for AI context

Your 200-page brand manual exceeds AI context windows. Critical rules get truncated or ignored entirely.

Structure gets lost

Do/Don't tables, color codes, logo placement rules – all the structured guidance disappears when AI reads raw text.

AI "forgets" rules mid-task

Even when rules are in context, AI agents drift. Without enforcement, your brand voice becomes inconsistent.

No audit trail

When AI makes a mistake, you can't trace which rule was violated or why. No accountability, no improvement.

What you get

A complete skill pack, ready for your AI agent runtime.

Corpus

Full text extraction as JSON with page references. Tables preserved. Language detected automatically.

Policies

MUST, SHOULD, MUST_NOT rules extracted via AI. Each rule linked to source page. Best results with English/German docs.

Skills & Validators

8 skill templates (LinkedIn, Press, Email, etc.). Automated validators for brand compliance checking.

No content is invented. If something can't be extracted, it's flagged – never fabricated.

Pricing

Pay per conversion. Price depends on document size. No subscriptions.

Available now

Pay per Use

Single conversion

From $5

price shown before payment

Upload one PDF
Full text & table extraction
Policy extraction (EN/DE)
8 skill templates
Automated validators
Download as JSON

Credit Pack

For multiple PDFs

Coming soon

bulk discount

Pre-purchase credits
Use anytime
Credits never expire
Volume discount
Priority processing
Email support

Enterprise

Self-hosted license

Custom

one-time purchase

Full source code
Run on your servers
Unlimited conversions
Custom integrations
No data leaves your infra
Dedicated support

How pricing works

Price is calculated based on your document: $5 base + $0.05 per page + $0.50 per OCR page (scanned pages). You see the exact price after upload, before payment. Example: A 100-page PDF costs approximately $10.

Live Processing

Watch your PDF transform

Upload → Analyze → Extract → Download

Drop your PDF here

or browse files

PDF files only, max 100 MB

Upload

Starting ingestion

Extracting pages

Detecting headers/footers

Chunking

Extracting tables

OCR (if needed)

Language detection

QA Gates

Creating pack

Install in 5 minutes

Unzip the pack, verify the manifest, load into your agent runtime.

# Install in Cursor IDE

1. Unzip the downloaded skill pack:
   unzip brand-skills-v1.0.0.zip

2. Copy to your Cursor skills folder:
   # macOS/Linux
   cp -r skills/* ~/.cursor/skills/
   
   # Windows
   xcopy skills %USERPROFILE%\.cursor\skills\ /E

3. Restart Cursor or reload window

4. The skill is now available!
   Cursor will automatically use the rules
   when working on related projects.

# Verify installation
ls ~/.cursor/skills/
# Should show your skill folder with SKILL.md

✓Enable validators as a gate before every generation

✓Use versioning for change tracking

✓Integrate tests into your CI/CD pipeline

Frequently Asked Questions

Text-based PDFs work best – Word exports, InDesign exports, and professionally created documents. Scanned PDFs are processed via OCR but quality depends on scan resolution.

We detect 50+ languages automatically. However, policy extraction (the AI-powered rule detection) is currently optimized for English and German documents. Other languages may work but with reduced accuracy.

Scanned PDFs are automatically detected and processed via OCR. You'll see in the live log which pages require OCR. OCR adds $0.50 per page to the price.

Your PDF is processed and then deleted. Generated skill packs are available for download for 7 days, then automatically deleted. We do not store your content permanently.

Text and table extraction is highly reliable for well-formatted PDFs. Policy extraction uses AI (Claude) which works well but should be reviewed. Every extracted rule includes a page reference so you can verify it.

A ZIP file containing: corpus.json (extracted text chunks), policies.json (MUST/SHOULD/MUST_NOT rules), skills.json (8 skill templates), validators.json (automated checks), and a manifest file.

Yes. All output is standard JSON format. You can load the policies and validators into any AI agent framework. We provide the structure – you integrate it into your runtime.

Agent Skills are structured rule packages for AI agents: policies, validators, and tests bundled together. Instead of hoping your AI remembers rules from a prompt, skills provide machine-readable policies with validators that check compliance automatically.

Roadmap

Converting legacy brand assets into AI-ready skill packs.

PDF Brand Manuals & Playbooks

Available

Extract text, tables, and structure from PDF documents. Works best with text-based PDFs. OCR available for scanned pages.

English & German

Available

Full policy extraction support for English and German documents.

ZIP Bundle Upload

Planned

Upload a ZIP containing your CD Manual (PDF/DOCX), logos (SVG/PNG), fonts, and playbooks. AI structures everything into a complete, ready-to-use Skill Package.

DOCX & Word Support

Planned

Read and extract policies from Microsoft Word documents and other office formats.

More Languages

Planned

Policy extraction for French, Italian, Spanish, Portuguese, and Russian.

Figma Design Systems

Planned

Import design tokens, component specs, and style guides directly from Figma files.

Sketch & Adobe XD

Future

Support for Sketch libraries and Adobe XD design systems.

Real-time Brand Monitoring

Future

Connect to your content pipeline and validate brand compliance in real-time.