Deck Analyzer Skill

AI-powered pitch deck analysis that extracts structured company information from various deck sources.

When to Use This Skill

Use this skill when you need to:

•Analyze a pitch deck from a URL or text
•Extract company information (name, product, team, traction)
•Prepare HubSpot descriptions from deck content
•Quick deck review without manual reading
•Find deck links in email text or messages

Natural language triggers:

•"analyze this deck: [URL]"
•"extract info from this pitch deck"
•"what's in this deck?"
•"review this pitch deck"
•"get company info from deck"

Features

•🔗 Multi-source support - DocSend, Google Drive, Dropbox, PDF links
•🤖 AI-powered extraction - Uses Claude to understand deck content
•📊 Structured output - Company, product, team, GTM, traction, fundraising
•💼 HubSpot-ready - Formats descriptions for CRM entry
•🔍 Link detection - Automatically finds deck links in text

Actions

analyze

Analyze a pitch deck from a URL and extract structured information.

Usage:

bash

deck-analyzer analyze <DECK_URL> [SENDER_EMAIL]

Arguments:

•DECK_URL - Link to the pitch deck (DocSend, Google Drive, Dropbox, PDF)
•SENDER_EMAIL - (Optional) Email of sender for DocSend authentication

Example:

bash

deck-analyzer analyze https://docsend.com/view/abc123xyz
deck-analyzer analyze https://docs.google.com/presentation/d/1abc/edit

Output:

json

{
  "company_name": "Acme Corp",
  "product_overview": "AI-powered workflow automation platform",
  "problem_solution": "Companies waste 40% time on manual tasks...",
  "key_capabilities": "Natural language automation, integration hub...",
  "team_background": "CEO: ex-Google AI, CTO: ex-Microsoft Azure...",
  "gtm_strategy": "Enterprise B2B, starting with Fortune 500...",
  "traction": "3 pilot customers, $50K ARR...",
  "fundraising": "Raising $2M seed round"
}

HubSpot-formatted description also provided

extract-links

Extract all deck links from text (email body, message, etc.).

Usage:

bash

deck-analyzer extract-links <TEXT_FILE>
# Or via stdin:
echo "Check out our deck: https://docsend.com/view/abc123" | deck-analyzer extract-links

Supported formats:

•DocSend: https://docsend.com/view/...
•Google Docs: https://docs.google.com/...
•Google Drive: https://drive.google.com/...
•Dropbox: https://www.dropbox.com/...
•PDF files: https://*/**.pdf

test

Run a test analysis with sample deck content.

Usage:

bash

deck-analyzer test

Tests the analyzer with a built-in sample deck to verify configuration.

Extracted Information

The analyzer extracts these fields:

•Company Name - Company or startup name
•Product Overview - 1-2 sentence summary of the product
•Problem/Solution - Problem being solved and the solution
•Key Capabilities - Main features or differentiators
•Team Background - Founders and key team members
•GTM Strategy - Go-to-market approach and target customers
•Traction/Validation - Current traction, customers, metrics
•Fundraising Ask - Funding amount and use of funds

If a field is not mentioned in the deck, it's omitted from output.

Environment Requirements

Required:

•ANTHROPIC_API_KEY - Claude API key for deck analysis

Optional:

•Sender email for DocSend authentication

Integration Examples

From OpenClaw Agent

bash

openclaw agent ask "Analyze this deck: https://docsend.com/view/abc123"

The agent will automatically use this skill to fetch and analyze the deck.

From Deal Logger

The deal-logger skill can automatically use deck-analyzer when it finds deck links in emails:

python

# In deal logging workflow
deck_links = extract_deck_links(email_body)
if deck_links:
    deck_info = analyze_deck(deck_links[0], sender_email)
    # Use deck_info to enrich deal description

Manual CLI

bash

# Analyze a specific deck
~/.openclaw/skills/deck-analyzer/deck-analyzer analyze https://docsend.com/view/abc123

# Extract links from an email
cat email.txt | ~/.openclaw/skills/deck-analyzer/deck-analyzer extract-links

# Test the analyzer
~/.openclaw/skills/deck-analyzer/deck-analyzer test

Technical Details

•Language: Python 3
•AI Model: Claude Haiku 4.5 (fast, cost-effective)
•Dependencies: requests (HTTP client)
•Timeout: 60 seconds per analysis
•Content limit: First 20,000 characters of deck

Output Format

JSON format for programmatic use:

json

{
  "company_name": "...",
  "product_overview": "...",
  ...
}

HubSpot format for CRM entry:

code

PRODUCT: AI-powered workflow automation platform

PROBLEM/SOLUTION: Companies waste 40% of time on manual tasks...

KEY CAPABILITIES: Natural language automation, integration hub...

TEAM: CEO: ex-Google AI, CTO: ex-Microsoft Azure

GTM STRATEGY: Enterprise B2B, starting with Fortune 500

TRACTION: 3 pilot customers, $50K ARR

FUNDRAISING: Raising $2M seed round

Use Cases

Quick Deck Review

Before a meeting, quickly extract key info from a deck:

bash

deck-analyzer analyze https://docsend.com/view/abc123

Automated Deal Enrichment

When deal-logger finds a deck in an email, automatically analyze it:

python

deck_info = analyze_deck(deck_url, sender_email)
company_description = format_company_description(deck_info)
# Add to HubSpot deal

Batch Deck Processing

Process multiple decks from a list:

bash

while read url; do
  deck-analyzer analyze "$url"
done < deck_urls.txt

Limitations

•DocSend authentication: Some DocSend links require email verification
•Content extraction: Only extracts visible text, not images
•PDF support: Basic PDF text extraction (may miss formatted content)
•Rate limits: Subject to Claude API rate limits

Troubleshooting

"ANTHROPIC_API_KEY not set"

bash

export ANTHROPIC_API_KEY="sk-ant-..."

DocSend access denied:

•Try providing sender email: deck-analyzer analyze URL sender@email.com
•DocSend may require browser-based authentication

No information extracted:

•Deck may be image-based (no text to extract)
•Content may be behind authentication
•Try downloading PDF and analyzing locally

Timeout errors:

•Large decks may take longer
•Try again or use a faster network connection

Privacy & Security

•Deck content is sent to Claude API for analysis
•No deck content is stored permanently
•Authentication cookies are not persisted
•Sender emails only used for DocSend auth requests