Where Are the Outputs from the patent-disclosure-skill Saved?

All outputs from the patent-disclosure-skill are written beneath the top-level outputs/ directory in the repository root, with each sub-skill creating dedicated sub-folders to organize generated artifacts.

The patent-disclosure-skill open-source repository follows a strict directory convention that keeps all generated files isolated under a single outputs/ tree. This design ensures that search reports, patent analyses, office action drafts, and CAD visualizations remain easy to locate while staying out of version control.

Patent Search Output Location

The patent search sub-skill writes its markdown search reports to outputs/patent-search/ by default. Files are named with timestamps, such as SEARCH-YYYYMMDD-HHMMSS.md, making chronological sorting automatic.

In skills/patent-search/tools/emit_search_report.py, lines 6-7 define the default output path relative to the repository root:

# From skills/patent-search/tools/emit_search_report.py

output_dir = Path("outputs/patent-search")

Patent Reader Output Location

The patent reader sub-skill stores downloaded PDFs, extracted text, intermediate JSONs, and final Obsidian notes beneath outputs/patent_reader/. When running multiple analyses, you can organize outputs into run-specific folders like outputs/patent_reader/<RUN>/.

According to skills/patent-reader/README.md (lines 21-24), the default PATENT_READER_OUTPUT_DIR environment variable or CLI argument points to this location, ensuring all artefacts from patent decoding workflows remain consolidated.

Patent Office Action (OA) Output Location

For office action workflows, generated opinion drafts, structured YAML, and final DOCX files land in outputs/oa/<case>/. The specific case folder organizes documents by patent application or matter.

The skills/patent-oa/tools/emit_opinion_docx.py script (lines 7-8) demonstrates this path construction, writing files such as outputs/oa/案/意见陈述_时间戳.md for Chinese patent cases.

Exam Policy Brief Output Location

Exam-policy briefs are saved to outputs/exam-policy/ with filenames following the pattern POLICY-YYYYMMDD-HHMM.md. This sub-skill generates markdown summaries of examination guidance and backlog statistics.

The prompt template in skills/patent-exam-policy/prompts/emit_backlog.md (lines 8-9) references this directory as the standard destination for policy artefacts.

Patent Disclosure CAD Output Location

The CAD/STEP processing sub-skill handles technical drawing generation, placing CAD views, SVG screenshots, and structure line-art in outputs/case/ or the recommended outputs/{案件}/cad_views/. You can override this with the --out-dir CLI flag.

In skills/patent-disclosure/tools/step_to_views.py, lines 475-476 establish the default:

# From skills/patent-disclosure/tools/step_to_views.py

default_out = f"outputs/{case_name}/cad_views/"

Output Directory Structure and Git Ignore

All outputs/ subdirectories are git-ignored by default via the repository's .gitignore file. This prevents generated artefacts from being committed upstream while keeping them locally available for downstream processing and review.

Practical Usage Examples

Run these commands to generate outputs in their default locations:

# Patent search report

python skills/patent-search/tools/cnipa_search.py \
    --inventor "John Doe" \
    --output-dir outputs/patent-search

# Patent reader extraction

python skills/patent-reader/tools/extract/fetch_patent_pdf.py \
    -o outputs/patent_reader/run1 \
    --pub CN12345678A

# Office action opinion draft generation

python skills/patent-oa/tools/emit_opinion_docx.py \
    -i outputs/oa/case/opinion_20260903120000.md

# CAD view generation from STEP files

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

Summary

• Root location: All artefacts reside under the repository's top-level outputs/ directory.
• Sub-skill isolation: Each tool creates its own sub-folder (patent-search/, patent_reader/, oa/, exam-policy/, case/) to prevent file collisions.
• Version control safety: The entire outputs/ tree is git-ignored by default to prevent accidental commits of generated files.
• Configurable paths: Most tools accept --output-dir or -o flags to override defaults while maintaining the conventional structure.

Frequently Asked Questions

Can I change the default output directory for patent-disclosure-skill?

Yes. While the outputs/ directory is the convention, most sub-skills accept CLI arguments such as --output-dir or -o to specify alternative paths. For example, step_to_views.py uses --out-dir to override the default outputs/case/ location, and the patent-reader supports arbitrary output paths via the -o flag.

Are output files automatically committed to git?

No. According to the repository's .gitignore configuration, all directories beneath outputs/ are excluded from version control by default. Generated reports, PDFs, and CAD files remain local to your working copy and will not be pushed upstream.

What file formats does the patent-disclosure-skill generate?

The skill produces diverse formats: markdown reports (SEARCH-YYYYMMDD-HHMMSS.md and POLICY-YYYYMMDD-HHMM.md), PDF documents, JSON extracts, YAML structures, DOCX opinion drafts, SVG screenshots, and CAD view files. Each sub-skill targets formats appropriate for its specific patent workflow, with source code in files like emit_search_report.py and emit_opinion_docx.py controlling the specific extensions.

How does the patent-reader organize multiple extraction runs?

The patent-reader supports run-specific sub-folders under outputs/patent_reader/<RUN>/, allowing you to maintain separate working directories for different patents or analysis sessions without overwriting previous extractions. This pattern is documented in skills/patent-reader/README.md and enables parallel processing of multiple patent publications.

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