How to Initiate Patent Point Excavation in Mode A: Complete Workflow Guide

To initiate patent point excavation in Mode A, invoke the skill with a disclosure request (e.g., /交底书 发明 项目路径 /my/project), and the system automatically executes excavation as the first logical step of the workflow.

Patent point excavation is the foundational stage of the disclosure workflow in the handsomestWei/patent-disclosure-skill repository. When operating in Mode A, the system automatically initiates patent point excavation to identify novel technical aspects from your project materials before generating any disclosure documents. This guide explains the exact commands, source files, and internal logic that drive this automatic process.

What Is Mode A in Patent Disclosure?

Mode A represents the patent-disclosure workflow (交底书主流程)—the core process that transforms technical project materials into structured disclosure documents. According to the repository's SKILL.md file at line 151, Mode A is defined in the section "模式 A · 交底书主流程" and automatically enters the "专利点挖掘" (patent-point excavation) stage immediately upon invocation. This stage identifies candidate invention points, utility model points, or design points that can serve as claim subject-matter in subsequent patent drafting.

How Patent Point Excavation Starts Automatically

Unlike other modes that require manual triggering of analysis steps, patent point excavation requires no explicit command. When you request the skill to "写交底书" (write disclosure) or use the /交底书 command, the system invokes Mode A and treats excavation as the mandatory first step. The workflow proceeds through intake, project scanning, and point extraction without additional user intervention until candidate points are generated.

Core Components of the Excavation Architecture

The excavation stage relies on three integrated components defined in the source code:

Intake and Mode Selection (tools/shared/run_step_to_views.py)

The entry point resides in tools/shared/run_step_to_views.py, which parses natural-language commands like /交底书 发明 项目路径 /home/user/my_project. This module determines that Mode A is required and initializes the workflow sequence.

Project Scanning (tools/shared/project_scan.md)

The project scanner walks the supplied directory, extracts text from source files, .docx, .pptx, and optional CAD files, then compiles a comprehensive project_scan.md file. This markdown document serves as the raw material for subsequent analysis.

Patent Point Extraction (build_context_anchor.py and validate_public_clues.py)

The extraction logic resides in two key files:

• tools/patent_reader/analyze/build_context_anchor.py: Creates the context anchor by analyzing scanned material using the appropriate schema
• tools/patent_reader/analyze/validate_public_clues.py: Validates extracted points against prior-art constraints and public clues

These modules apply schema-specific rules from invention_schema.yaml, utility_model_schema.yaml, or design_schema.yaml to identify patentable technical features.

Step-by-Step Execution Flow

When you initiate Mode A, the skill executes this exact sequence:

1. User Invocation – Submit a command like /交底书 发明 项目路径 /my/project
2. Intake Processing – The system gathers patent type and project location via the intake prompt
3. Project Scanning – The scanner collects textual artifacts and generates tools/shared/project_scan.md
4. Point Extraction – The extractor analyzes the scan using the appropriate schema and produces candidate lists
5. Output Generation – Results are stored in outputs/patent_points/ as invention_points.yaml, utility_model_points.yaml, or design_points.yaml

Command Examples to Trigger Excavation

You can initiate the process using natural language or explicit flags:

# Natural language invocation (recommended)

/交底书 发明 项目路径 /home/user/my_project

# Explicit mode flag (alternative syntax)

/patent-disclosure-skill --mode A --type invention --path /home/user/my_project

Once received, the skill executes internal calls through the following logic (illustrative):

# Entry point in tools/shared/run_step_to_views.py

run_step_to_views.handle_user_input(user_input)

# Mode A sequence selection

if mode == "A":
    steps = [
        "intake",
        "project_scan", 
        "patent_point_excavation",
        "prior_art_search",
        "figure_generation",
        "disclosure_builder",
    ]

# Patent point extraction in tools/patent_reader/analyze/build_context_anchor.py

anchor = build_context_anchor.extract_points(project_scan_path, schema_path)

Accessing Excavated Patent Points

After extraction completes, inspect the generated point lists directly:

# View invention points

cat outputs/patent_points/invention_points.yaml

# View utility model points  

cat outputs/patent_points/utility_model_points.yaml

Each file contains candidate points validated by validate_public_clues.py and structured according to the respective schema requirements.

Summary

• Patent point excavation in Mode A starts automatically when you invoke the /交底书 command—no separate trigger is required
• The workflow is defined in SKILL.md (line 151) and orchestrated through tools/shared/run_step_to_views.py
• Project scanning generates project_scan.md, which feeds into extractors in tools/patent_reader/analyze/
• Output files (invention_points.yaml, etc.) are stored in outputs/patent_points/ for user confirmation before proceeding to prior-art search and disclosure generation

Frequently Asked Questions

Do I need a separate command to start patent point excavation?

No. Patent point excavation is the first logical step of Mode A and executes automatically after you provide a disclosure request. The skill handles the transition from intake to excavation without requiring manual intervention.

What file types does the project scanner analyze?

The scanner processes source code files, .docx documents, .pptx presentations, and optional CAD files. It aggregates all textual artifacts into tools/shared/project_scan.md for the extraction engine to analyze.

Where are the extracted patent points stored?

Candidate points are saved as YAML files in the outputs/patent_points/ directory. The specific filename depends on the patent type selected during intake (e.g., invention_points.yaml for invention patents).

Which source files handle the validation of extracted points?

Validation occurs in tools/patent_reader/analyze/validate_public_clues.py, which checks candidate points against public clues and prior-art constraints to ensure they meet patentability requirements before presenting them to the user.

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