Skip to content

Internal Automation Tools

Zero-Shot Agency relies on custom Python scripts to track Prompt Share of Voice (SOV) and generate bot-native assets. These tools automate the collection and publishing mechanics required for Generative Engine Optimization (GEO).

1. GEO Tracker (geo-tracker.py)

A script to monitor major LLMs (OpenAI, Anthropic, Gemini) and track how frequently they mention or cite a specific target domain across a set of target queries. This gives us our "Share of Voice" metric.

Requirements

pip install openai anthropic google-generativeai

Set the required API keys in your environment: - OPENAI_API_KEY (for GPT-4o) - ANTHROPIC_API_KEY (for Claude 3.7 Sonnet) - GEMINI_API_KEY (for Gemini 1.5 Pro)

Note: If an API key is missing, the script will skip querying that specific model and print a warning.

Usage

python geo-tracker.py --domain example.com --queries "query 1" "query 2" "query 3"

Arguments

  • --domain (Required): The target domain to track (e.g., 'example.com').
  • --queries (Required): A space-separated list of queries to test.

Output

The script performs synchronous queries (with a 1-second delay between queries to respect rate limits) against all configured LLM APIs. It then prints a Share of Voice report detailing the percentage of times the target domain was mentioned in the output for each engine.

2. LLMs.txt Generator (llms-txt-generator.py)

A specialized web crawler that traverses a domain and builds a standardized /llms.txt and llms-full.txt file. This is crucial for formatting site content correctly for LLM and bot consumption.

Requirements

pip install requests beautifulsoup4 markdownify

Usage

python tools/llms-txt-generator.py https://example.com --max-pages 50 --output ./site

Arguments

  • url (Required): Starting URL to crawl (e.g., https://example.com).
  • --max-pages: Maximum number of pages to crawl (Default: 50).
  • --output: Output directory for the generated files (Default: current directory).
  • --delay: Delay between requests in seconds to respect server loads (Default: 0.5).

Output

  • llms.txt: A markdown index file containing links to all crawled pages.
  • llms-full.txt: A concatenated markdown file containing the parsed and cleaned text of all main content across the crawled domain.

3. Publisher Pipeline (publisher_pipeline.py)

An autonomous script that takes an approved markdown draft and publishes it across multiple platforms simultaneously. It pushes to GitHub, logically splits the content into a Twitter/X thread (<280 chars per tweet), and emails the full draft via Substack.

Requirements

Ensure the following CLI tools are installed and available in your PATH: - git - x-cli (for Twitter) - himalaya (for Substack email)

Set the required environment variable: - SUBSTACK_EMAIL (the destination email address for your Substack drafts)

Usage

python tools/publisher_pipeline.py path/to/draft.md

Arguments

  • filepath (Required): Path to the markdown file you want to publish.

Output

The script extracts the title from the markdown file, commits and pushes it to the main branch via git, uses x-cli to post a threaded summary to X, and pipes the full content through himalaya to send it to the configured Substack email address.

(Source code is maintained in the tools/ and docs/tools/ directory of our repository).