How to Trigger the Patent Disclosure Sub-Skill in Instagit Environments

You cannot trigger the patent-disclosure sub-skill directly because it is explicitly configured as non-user-invocable (user-invocable: false in SKILL.md), so you must invoke the parent skill via API call (await agent.run_skill("patent-disclosure")) or chat command (开始专利交底书) to execute the workflow.

The patent-disclosure package in the handsomestWei/patent-disclosure-skill repository is designed as an internal sub-skill that orchestrates the creation of patent disclosure documents. Unlike standalone skills, this component is explicitly restricted from direct user invocation and must be triggered through a higher-level parent skill that manages the complete workflow.

Why Direct Invocation Is Blocked

In skills/patent-disclosure/SKILL.md, the skill declares user-invocable: false, which explicitly blocks end-users from activating it with direct commands like !patent-disclosure. This architectural choice ensures the multi-step patent disclosure process—spanning invention mining, prior-art search, and document generation—executes in the correct sequence under parent skill supervision rather than as isolated calls.

Methods to Trigger the Patent Disclosure Workflow

To execute the sub-skill, you must interact with the parent skill using one of the following supported methods.

Programmatic Invocation via Instagit SDK

When building automation or integrating with the Instagit agent framework, call the parent skill programmatically. The parent skill loads the patent-disclosure package and runs the step-by-step prompts defined in its prompts/ directory:

from instagit import Agent

agent = Agent()

# Triggers the parent skill, which internally runs the patent-disclosure sub-skill

await agent.run_skill("patent-disclosure")

Natural Language Chat Trigger

In chat-enabled environments, type the Chinese command that maps to the parent skill:

开始专利交底书

The platform recognizes this intent, launches the parent skill, which then instantiates the sub-skill and begins executing the sequence of prompts located in skills/patent-disclosure/prompts/.

Execution Flow After Triggering

Once triggered, the parent skill loads the package and executes the comprehensive workflow. The process includes:

• Intake and project scanning: Collecting invention details and analyzing relevant codebase sections
• Mining phases: Extracting utility model and design elements through specialized prompts
• Prior-art search: Executing tools/cnipa_epub_search.py for lightweight Chinese patent database queries
• Document construction: Using tools/md_to_docx.py and tools/browser.py to generate and format final Word documents
• Self-check and iteration: Running validation steps defined in the prompt files

Critical Source Files

Understanding these repository paths clarifies the trigger mechanism and restrictions:

• skills/patent-disclosure/SKILL.md: Contains the user-invocable: false flag and defines the execution plan that the parent skill follows
• skills/patent-disclosure/prompts/: Directory containing markdown files for each workflow step (intake, mining, search, preview, builder)
• skills/patent-disclosure/tools/: Utility scripts including cnipa_epub_search.py, browser.py, and md_to_docx.py used during execution
• INSTALL.md: Lists prerequisites like Playwright and requirements-cnipa.txt required before the sub-skill can function
• README.md: Provides top-level usage instructions for the skill package

Summary

• The patent-disclosure sub-skill is explicitly marked as non-invocable in SKILL.md to prevent direct user access and ensure proper workflow sequencing
• Triggering requires invoking the parent skill either through the Instagit SDK (agent.run_skill) or the chat command 开始专利交底书
• Upon triggering, the sub-skill executes a multi-step workflow using prompts and tools from its respective directories under skills/patent-disclosure/
• Direct invocation attempts (e.g., !patent-disclosure) are ignored or rejected by the agent framework as implemented in handsomestWei/patent-disclosure-skill

Frequently Asked Questions

Can I trigger the patent-disclosure sub-skill directly from a user message?

No. The skill configuration in SKILL.md explicitly sets user-invocable: false, which prevents direct activation. Any attempt to call !patent-disclosure from a user message will be rejected by the agent framework because the skill is designed strictly for internal orchestration by parent skills.

What is the correct chat command to start the patent disclosure process?

Type 开始专利交底书 (meaning "Start patent disclosure document") in the chat interface. The platform maps this natural language intent to the parent skill, which then loads and executes the patent-disclosure sub-skill workflow defined in the repository's skills/patent-disclosure/ directory.

Does the sub-skill require specific dependencies before triggering?

Yes. According to INSTALL.md in the repository root, the environment must have Playwright installed and Python dependencies from requirements-cnipa.txt satisfied before the parent skill can successfully execute the sub-skill's tool scripts like cnipa_epub_search.py or md_to_docx.py.

What happens if I try to call agent.run_skill("patent-disclosure") without proper parent skill configuration?

The call will fail or return an error because patent-disclosure is registered as a sub-skill with restricted invocation rights, not a standalone skill. The agent requires the parent skill wrapper to handle the execution context and sequential prompt loading from skills/patent-disclosure/prompts/ that the sub-skill cannot manage independently.

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