yc-batch-evaluator

What it does

Scrapes a YC batch directory to enumerate companies, then researchers each company and its founders (via YC pages and Apollo), evaluates founder–company fit, analyzes the company website, and exports results to Google Sheets with a 12-column row per company that updates live as results complete. The process is designed to create a new Google Sheet for each run and to stream updates row-by-row in batches of 3-5 companies.

How it works

1. Setup credentials from ~/.gooseworks/credentials.json and export GOOSEWORKS_API_KEY and GOOSEWORKS_API_BASE. If credentials are missing, run npx gooseworks login.
2. Step 1: Scrape the YC Batch Directory by posting to the Goose API with api scrapegraph and path /v1/smartscraper, supplying the batch page URL. The system expects structured data of companies with fields: name, description, tags, location, and url_slug. If sectors are not provided at batch level, the pages may provide richer sectors data later.
3. Step 2: Create a new Google Sheet via an API call to google-sheets with path /create-spreadsheet, then populate the header and initial rows with company names and descriptions. The sheet columns A-N are predefined, with A: Company, B: Description, C: Sector, D: Location, E: Website, F: Founders, G: Founder LinkedIn(s), H: Founder Twitter/X, I: Founder Background, J: Founder-Company Fit, K: Website Analysis, L: Market/Competitors, M: Overall Assessment, N: Priority Rank.
4. Step 3: Research Each Company — Row by Row
   • 3a. Scrape YC company page to obtain full description, founders (with LinkedIn and Twitter URLs), website_url, location, sectors, team_size, founding_year. Exclude YC staff from founders. Use fields like website_url, sectors, and location from the page data, preferring sectors from the individual page over batch tags if available.
   • 3b. Scrape the company’s own website for product, pricing, features, traction signals, and hiring signals; return fields such as product_description, target_customer, pricing, key_features, traction_signals, hiring_signals.
   • 3c. Apollo — founder work history: call with founder LinkedIn URL to fetch employment_history, city/state for location fallback, and headline. If LinkedIn is missing, fallback to name + company via first_name, last_name, and organization_name.
   • 3d. Perplexity — market context: provide market size/opportunity, top competitors, founder qualification, traction or mentions, and red flags.
5. Step 4: Compile and Update Each Row
   • For each completed company, perform a Google Sheets update with exactly 12 values in this order: sectors, location, website, founders, linkedins, twitters, background, fit, website_analysis, market, overall, (blank for N). Each row is updated via its own update-values call as soon as the data is ready, not batched.
6. Formatting rules
   • All links must be plain URLs (no HYPERLINK formatting).

When to use it

Use when asked to evaluate YC companies, research a YC batch, screen startups, or do due diligence on YC companies. The skill defaults to Spring 2026 batch if not specified and aims to process all companies when sector/thesis filters are not provided.

What it can touch

The workflow references endpoints for: scrapegraph, google-sheets (create-spreadsheet and update-values), apollo/people/match, and perplexity. It requires an API key and base URL from ~/.gooseworks/credentials.json and utilizes Bearer authentication.

Caveats

The tool requires explicitly starting the process without clarifying questions. It creates a new Google Sheet for each run and streams updates in near real-time. The batch page may provide only high-level tags; individual YC pages provide richer sectors data. Some fields (LinkedIn, Twitter) may be null and are handled with fallbacks.