Requirements for Generating Patent Application Documents with the Patent-Application Skill

To generate patent application documents with the patent-application skill, you must explicitly invoke the skill by name, provide a disclosure directory containing three mandatory artifacts (交底书, schema file, and line drawings), and select the patent type (invention, utility model, or design).

The patent-application skill in the handsomestWei/patent-disclosure-skill repository transforms prepared technical disclosures into complete patent-application packages, including claims, specifications, abstracts, and drawings. Understanding the strict requirements encoded in SKILL.md and enforced by the validation tools ensures successful document generation without runtime failures.

Prerequisites and Pre-Conditions

The skill enforces a rigorous set of pre-conditions before entering the document generation pipeline. These requirements are defined in SKILL.md and validated by tools/material_gate.py.

Mandatory Artifact Requirements

The disclosure directory must contain three specific files. The skill cannot fabricate missing data and aborts with exit code 2 if any are absent, prompting you to run the patent-disclosure skill first:

  • Disclosure document (交底书.md) – The core technical disclosure content
  • Schema file (*.schema.yaml) – Structured metadata defining the technical fields
  • Line-drawing(s) (line_drawings/) – Black-and-white technical diagrams required for the application

According to the source code in SKILL.md (lines 9-10), the material gate explicitly checks for these artifacts before proceeding.

Explicit User Invocation and Patent Type Selection

You must explicitly name the skill during invocation (e.g., "申请文件" or "patent-application") to prevent accidental execution.

Additionally, you must specify the patent type, which determines the downstream pipeline:

Execution Workflow and Pipeline

The skill follows a deterministic workflow defined in SKILL.md, ensuring reproducible outputs and proper error handling.

Material Gate Validation

Before any document generation, the skill runs tools/material_gate.py to enforce isolation and verify inputs:

python skills/patent-application/tools/material_gate.py \
    --case-dir outputs/<case-id>

If validation fails, the script exits with code 2, requiring you to complete the disclosure phase first.

Document Generation Pipeline

For invention and utility model patents, the skill orchestrates a multi-stage pipeline:

  1. Guardrails & Intake – Processes prompts/guardrails.md → intake.md
  2. Iteration Handling – If revising existing output, reads iteration_context.md and creates a new timestamped folder
  3. Claim Strategy – Executes claim_strategy.md → claims_builder.md
  4. Figure Processing – Runs figures.md and numeral_register.md
  5. Specification Building – Generates technical content via specification_builder.md
  6. Consistency Checks – Validates cross-references through consistency.md
  7. Design Pipeline – For design patents, only design_application.md executes

Output Structure and Versioning

Results are written to immutable, timestamped directories to prevent overwrites and maintain audit trails:


outputs/patent-application/{case-id}_{timestamp}/

As implemented in SKILL.md (lines 15-16, 21-22), each run generates a new folder rather than modifying existing outputs, supporting proper versioning.

Command Examples for Generating Patent Documents

Below are practical commands that satisfy all requirements for generating complete patent application packages.

Validate Disclosure Directory Structure

Ensure your disclosure directory contains the required artifacts before invocation:

tree outputs/案件A

# Expected layout:

# ├─ 交底书.md

# ├─ schema.yaml

# └─ line_drawings/

#     └─ fig1.png

Run Material Gate Validation

Verify all prerequisites are met before generating documents:

python skills/patent-application/tools/material_gate.py \
    --case-dir outputs/案件A

# Exit code 0 → all artifacts present

# Exit code 2 → missing required files → run patent-disclosure first

Generate Full Application Package (Invention)

Execute the complete pipeline and emit the Word document:

python skills/patent-application/tools/emit_application_docx.py \
    --dir outputs/patent-application/案件A_$(date +%Y%m%d%H%M%S)

This command orchestrates the full workflow, performs consistency checks, renders the .docx file, and produces 问题清单.md (problem list) if ambiguities arise.

Generate Design Application Only

For design patents, the skill internally invokes the design-specific pipeline:

python skills/patent-application/tools/design_application.py \
    --case-dir outputs/案件B

Summary

  • Three mandatory artifacts (交底书, schema, line drawings) must exist in the disclosure directory, or the skill exits with code 2.
  • Explicit invocation by skill name and patent type selection are required to trigger the appropriate pipeline.
  • Material gate validation (tools/material_gate.py) enforces pre-conditions before document generation begins.
  • Timestamped output folders ensure immutable, versioned results in outputs/patent-application/.
  • Problem list generation (问题清单.md) captures content ambiguities without blocking the main workflow.

Frequently Asked Questions

What happens if the disclosure directory is missing required files?

The skill aborts immediately. tools/material_gate.py checks for the presence of 交底书.md, schema.yaml, and line drawings, exiting with code 2 if any are missing. You must run the patent-disclosure skill first to generate these prerequisite artifacts.

Can I overwrite an existing patent application output?

No. According to SKILL.md lines 15-16, the skill enforces immutability by creating a new timestamped folder for each run. If you are revising an existing application, the skill reads iteration_context.md and generates a fresh output directory rather than overwriting previous results.

What is the difference between invention and design patent generation workflows?

Invention and utility model patents trigger a comprehensive pipeline (claim_strategy.md → claims_builder.md → figures.md → specification_builder.md → numeral_register.md → consistency.md), while design patents execute only design_application.md. The patent type selection in SKILL.md (lines 17-18) determines which sub-pipeline runs.

Where are content issues or ambiguities recorded during generation?

Any content-level disputes are recorded in 问题清单.md within the output directory. As specified in SKILL.md (lines 10-11), the skill surfaces the path to this problem list along with a short summary at the end of the dialogue, allowing you to address issues without interrupting document generation.

Have a question about this repo?

These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:

Share the following with your agent to get started:
curl -s "https://instagit.com/install.md"

Works with
Claude Codex Cursor VS Code OpenClaw Any MCP Client

Maintain an open-source project? Get it listed too →