Output Paths for Different Sub-Skills in the Patent Disclosure Skill Repository

The handsomestWei/patent-disclosure-skill repository writes all generated artifacts to a top-level outputs/ directory, with dedicated sub-folders for Patent Search, Patent Reader, Patent OA, Exam-Policy, and case-specific Patent Disclosure folders.

Understanding the output paths for different sub-skills is essential for automating patent workflows and managing generated documents. In this repository, each sub-skill follows a strict convention that separates user-generated data from source code, ensuring clean version control and predictable file locations. This guide maps the exact directory structure as defined in the source code and SKILL files.

Output Directory Structure by Sub-Skill

Patent Search

Patent Search generates timestamped markdown reports in outputs/patent-search/SEARCH-YYYYMMDD-HHMMSS.md. As documented in skills/patent-search/SKILL.md lines 33-35, and enforced by emit_search_report.py lines 6-7, this sub-skill automatically creates the output directory and writes both markdown reports and optional JSON metadata files.

Patent Reader

Patent Reader organizes outputs under outputs/patent_reader/ or run-specific subdirectories like outputs/patent_reader/${RUN}/. According to skills/patent-reader/SKILL.md line 14 and tools like fetch_patent_pdf.py line 13, this structure stores PDF source files, extracted text, claim_tree.json, claim_mermaid.mmd diagrams, and final observation markdown files.

Patent OA (Office Action)

Patent OA uses case-specific folders located at outputs/oa/{案件}/, where {案件} represents the specific case identifier. Defined in oa/prompts/respond_office_action.md lines 107-135 and utilized by emit_opinion_docx.py line 7, these directories contain draft opinion markdown (意见陈述_时间戳.md), final Word documents, and structured YAML files such as notice_struct.yaml.

Exam-Policy

Exam-Policy generates policy briefs at outputs/exam-policy/POLICY-YYYYMMDD-HHMM.md. The path convention is documented in skills/exam-policy/SKILL.md line 14 and the emit_backlog.md prompt at line 8, storing policy brief markdown and optional appendix files with timestamped filenames.

Patent Disclosure

Patent Disclosure targets case-based directories such as outputs/case/ or outputs/{案件}/cad_views/. Tools located in skills/patent-disclosure/tools/—specifically run_step_to_views.py at line 475—use the --out-dir flag to write CAD screenshots, line-art PDFs, mermaid diagrams, generated disclosure markdown, and Word documents.

How Output Paths Are Defined in the Source Code

The output conventions are hard-coded in individual SKILL.md files and reinforced by command-line interfaces. The top-level SKILL.md establishes that all sub-skills write artifacts outside the skill package into the repository's top-level outputs/ folder. This design keeps user-generated data isolated from the codebase and respects the repository's .gitignore rules.

Most tools expose an -o/--output-dir or --out-dir parameter to override defaults. For example, skills/patent-disclosure/tools/run_step_to_views.py suggests the path "输出目录（建议 outputs/{案件}/cad_views/）" in its help text, while skills/patent-search/tools/emit_search_report.py defaults to outputs/patent-search/ when no alternative is provided.

Practical Examples for Each Sub-Skill

Running a Patent Search

python skills/patent-search/tools/cnipa_search.py \
    --query "发明人:张三" \
    --output-dir outputs/patent-search/

Result: outputs/patent-search/SEARCH-20231128-153200.md

Processing a Patent PDF

python skills/patent-reader/tools/extract/fetch_patent_pdf.py \
    -i CN119961390A.pdf \
    -o outputs/patent_reader/run1

Results:

• outputs/patent_reader/run1/source/CN119961390A.pdf
• outputs/patent_reader/run1/claim_tree.json
• outputs/patent_reader/run1/claim_mermaid.mmd

Drafting an Office Action Response

python skills/patent-oa/tools/emit_opinion_docx.py \
    -i outputs/oa/案1/意见陈述_20231201_1015.md \
    -o outputs/oa/案1/

Result: outputs/oa/案1/意见陈述_20231201_1015.docx

Generating CAD Views for Disclosure

python skills/patent-disclosure/tools/run_step_to_views.py \
    -i model.step \
    -o outputs/case/cad_views/

Result: Files saved under outputs/case/cad_views/

Creating an Exam-Policy Brief

python -c "import datetime; print(datetime.datetime.now().strftime('outputs/exam-policy/POLICY-%Y%m%d-%H%M.md'))"

Result: outputs/exam-policy/POLICY-20231201-1015.md

Summary

• All sub-skills write to outputs/ in the repository root, keeping user data separate from source code and respecting .gitignore rules.
• Patent Search uses auto-generated timestamped files in outputs/patent-search/ as defined in emit_search_report.py.
• Patent Reader organizes runs into subdirectories under outputs/patent_reader/ with files extracted by fetch_patent_pdf.py.
• Patent OA uses case-specific folders in outputs/oa/{案件}/ for draft opinions and final Word documents.
• Exam-Policy generates timestamped briefs in outputs/exam-policy/ according to exam-policy/SKILL.md.
• Patent Disclosure targets case folders like outputs/case/ or outputs/{案件}/cad_views/ via tools in patent-disclosure/tools/.
• Every tool supports -o/--output-dir or --out-dir flags to override default paths.

Frequently Asked Questions

Where does Patent Search save its reports by default?

By default, Patent Search saves reports to outputs/patent-search/SEARCH-YYYYMMDD-HHMMSS.md. This path is defined in skills/patent-search/SKILL.md and enforced by emit_search_report.py when no custom output directory is specified via the --output-dir flag.

How do I specify a custom output directory for Patent Reader?

Use the -o flag when running fetch_patent_pdf.py or other reader tools. For example, the command -o outputs/patent_reader/run1 creates a dedicated run folder containing the source PDF, extracted claim_tree.json, and mermaid diagrams, isolating each processing run.

Are output directories automatically created if they don't exist?

Yes. The tools in handsomestWei/patent-disclosure-skill automatically create the output directory structure, including case-specific subfolders like outputs/oa/{案件}/ or outputs/patent_reader/${RUN}/. The codebase ensures parent directories exist before writing files, preventing path errors.

How do I prevent output files from being tracked by Git?

The repository includes .gitignore rules that exclude the top-level outputs/ directory. By writing all artifacts to this location rather than inside skill packages, the design ensures that generated content remains untracked while the codebase stays clean and version-controlled.

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