Prerequisites for Running patent-disclosure-skill: Complete Setup Guide

To run patent-disclosure-skill successfully, you need Python 3.9 or newer, pip, a Chromium-based browser (Chrome, Edge, or Playwright-installed Chromium), and optionally separate virtual environments for CAD STEP parsing and Obsidian vault integration.

The patent-disclosure-skill is an AgentSkills-based automation framework designed for patent drafting, prior-art search, and Office Action (OA) assistance. Before invoking the skill from hosts like Claude or Cursor via SKILL.md, you must configure a Python-centric environment with specific browser dependencies and optional tooling extensions as documented in INSTALL.md.

Core Runtime Requirements

The foundation of the skill is a standard Python installation with package management capabilities.

• Python 3.9+ (64-bit): All scripts, prompts, and skill orchestration logic are pure Python. The runtime requires version 3.9 or higher according to INSTALL.md.
• pip: Required to install dependencies from the various requirements.txt files located throughout the repository.

Essential Python Dependencies

The skill uses multiple requirement files to separate core functionality from optional tools. You must install at least the core dependencies, then add others based on your use case.

Core dependencies (requirements.txt):

pip install -r requirements.txt

This installs critical packages including python-docx for Word document generation, playwright for browser automation, and latex2mathml for formula rendering.

Optional tool-specific dependencies:

# For CNIPA EPUB prior-art search

pip install -r tools/crawl/requirements-cnipa.txt

# For Office Action (OA) case handling and RAG workflows

pip install -r tools/oa/requirements-oa.txt

# For PDF patent extraction

pip install -r tools/patent_reader/requirements.txt

Browser Engine and Playwright Configuration

The skill requires a Chromium-based browser for rendering Mermaid diagrams, extracting PDF content, and performing CNIPA-EPUB searches. Without a browser, the skill can still generate Markdown, but images and Word document diagrams will be missing.

Option 1: Use existing system browsers (Google Chrome or Microsoft Edge installed on the host).

Option 2: Install Chromium via Playwright:

python -m playwright install chromium

Verify your setup using the browser probe utility:

python tools/shared/browser.py --probe

This command reports whether Chrome, Edge, or Playwright Chromium is available. Run the install command above only if the probe returns ok=false.

Optional Advanced Components

CAD STEP File Processing

To enable ingestion of .step or .stp files for utility-model patents (producing multi-view isometric projections), you need a separate virtual environment with CadQuery (Python 3.10–3.12) plus system Cairo libraries.

Run the bootstrap scripts:

python tools/shared/cad_venv.py
python tools/shared/bootstrap_cad_venv.py

Obsidian Vault Integration

For the "Patent Reader" mode that provides rich knowledge-graph and Canvas features, you need a local Obsidian vault with optional community plugins like Dataview.

Set the environment variable and verify:

export PATENT_READER_OBSIDIAN_VAULT="$HOME/Obsidian/MyVault"
python tools/patent_reader/vault/check_obsidian_env.py --auto-accept

Without Obsidian, the skill falls back to writing output to the outputs/patent_reader/ folder.

OA Vector Store and Embeddings

For the RAG-style "审查答复" (OA) workflow that retrieves similar past cases, configure an API key in tools/oa/config.py. Supported services include ZhipuAI, DashScope, OpenAI, and MiniMax, or you can use a local embedding model.

Step-by-Step Installation Workflow

Follow this sequence to prepare the environment:

# 1️⃣ Clone the repo into your skills folder

git clone https://github.com/handsomestWei/patent-disclosure-skill.git ~/.cursor/skills/patent-disclosure-skill

# 2️⃣ Install core Python dependencies

pip install -r requirements.txt

# 3️⃣ Verify browser availability; install Chromium if needed

python tools/shared/browser.py --probe
python -m playwright install chromium  # Only if probe reports no browser

# 4️⃣ (Optional) Install CNIPA crawler for prior-art search

pip install -r tools/crawl/requirements-cnipa.txt

# 5️⃣ (Optional) Setup CAD environment for STEP files

python tools/shared/cad_venv.py
python tools/shared/bootstrap_cad_venv.py

# 6️⃣ (Optional) Configure Obsidian vault path

export PATENT_READER_OBSIDIAN_VAULT="$HOME/Obsidian/MyVault"

Once satisfied, invoke the skill via your AgentSkills host using the entry point defined in SKILL.md. The skill automatically locates prompts, executes Python tools from the tools/ directory, and outputs artifacts to outputs/ or directly into your Obsidian vault.

Summary

• Python 3.9+ and pip are mandatory for all skill operations.
• Chromium-based browser (Chrome, Edge, or Playwright Chromium) is required for diagram rendering and PDF extraction; verify with tools/shared/browser.py.
• Core dependencies live in requirements.txt, while specialized tools have their own requirement files in tools/crawl/, tools/oa/, and tools/patent_reader/.
• Optional CAD support requires Python 3.10–3.12 in an isolated virtual environment managed by tools/shared/cad_venv.py.
• Optional Obsidian integration requires the PATENT_READER_OBSIDIAN_VAULT environment variable and the check script tools/patent_reader/vault/check_obsidian_env.py.
• OA RAG functionality requires embedding service configuration in tools/oa/config.py.

Frequently Asked Questions

Can I run patent-disclosure-skill without installing a browser?

Yes, but with degraded functionality. Without Chrome, Edge, or Playwright Chromium, the skill cannot render Mermaid diagrams to PNG, extract content from certain PDFs, or perform CNIPA-EPUB searches. It will still generate Markdown text and Word documents, but visual diagrams will be missing according to the INSTALL.md documentation.

What Python version is required for CAD STEP file processing?

The CAD STEP parsing feature requires Python 3.10 to 3.12 specifically, distinct from the core Python 3.9+ requirement. This runs in an isolated virtual environment created by tools/shared/cad_venv.py and initialized via tools/shared/bootstrap_cad_venv.py to avoid dependency conflicts with the main skill environment.

How do I verify my browser setup before running the skill?

Execute the probe command python tools/shared/browser.py --probe from the repository root. This script checks for system Chrome or Edge installations and Playwright status. If it reports ok=false, install the fallback with python -m playwright install chromium to ensure headless browser automation functions correctly for diagram rendering and crawling tasks.

Where do I configure API keys for the OA RAG feature?

Embedding service API keys (ZhipuAI, DashScope, OpenAI, or MiniMax) are configured in tools/oa/config.py. This configuration powers the vector-store similarity search for "审查答复" (Office Action) cases, enabling the skill to retrieve relevant historical responses when processing new OA workflows.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}