Paper to Chinese Patent Draft
Turn scientific papers, reports, and code into evidence-backed Chinese patent drafts and technical disclosures.
Evidence shows the skill mandates stable source IDs, marks unsupported features as 'unsupported' and excludes them from formal claims, does not infer inventorship/ownership/legal sufficiency, and asks users to confirm missing facts. These reflect data-flow transparency and user confirmation. However, no explicit least-privilege permissions, rollback, or external-effect control is documented, and the publisher is unverified. Deductions applied.
The skill defines a validation workflow (validate_patent_draft.py and build_patent_package.py) and has a tests directory with pytest in CI, but static review cannot execute these tests, and no explicit error-handling or failure-feedback design is visible. Cap at 10; given tests exist but no independent reproduction evidence, score 7.
The skill clearly defines use cases, typical requests, outputs, and boundaries, and supports Chinese outputs for Chinese users. However, core functionality may depend on overseas services (e.g., CNIPA search requiring Playwright and network), potentially unreachable from mainland China, and environment compatibility is not fully detailed. Score 10.
Documentation is well-structured with bilingual READMEs, follows information architecture largely, provides dependencies, examples, and boundaries. But hidden assumptions (Python dependencies, Playwright), no version history or changelog, and unclear maintenance responsibility (unverified publisher) lead to deduction.
The skill specifies detailed outputs (claims, specification, abstract, etc.) and an end-to-end process, but static review cannot verify output quality. Core task is clear and validation is specified, yet no execution evidence; cap at 7, score 6.
A CI workflow runs pytest for this skill, providing some third-party execution evidence, but test contents are not provided and independent reproduction is not possible. Cap at 5, score 4.
- The skill depends on overseas services (e.g., CNIPA search requiring Playwright and external network); access from mainland China may be restricted and core features may not be fully usable.
- Publisher is unverified; maintenance responsibility and update path are unclear. Verify recent commits and issue responsiveness before use.
- The output is a drafting aid for inventor and professional review, not a legal opinion; always have a patent attorney validate.
- The skill may execute scripts (e.g., validation, package building). Run in a controlled environment and review script behavior.
What does this skill do, and when should you use it?
This skill converts papers, theses, technical reports, source code, figures, inventor notes, or research manuscripts into Chinese invention patent drafts and attorney-facing technical disclosure materials. It first loads the workflow definition (manifest.yaml) and permanent files, detecting source format, task mode, and invention type. It then creates stable source IDs for paper text, equations, figures, and code, ensuring every claim feature maps to explicit evidence. The skill progresses through stage gates: building a source map, terminology ledger, and evidence inventory before drafting claims and specification. Final deliverables include Chinese DOCX files with native editable Office Math formulas and Mermaid diagrams.
Runs Python workflow scripts (validate_patent_draft.py, build_patent_package.py) to generate Chinese DOCX; extracts tables and formulas (embedded as native Office Math); renders flowcharts and system diagrams via Mermaid; performs internal consistency checks and produces validation reports; outputs claims, specification, abstract, figures, and technical disclosure in Markdown and DOCX.
- A researcher converts a published paper into a Chinese patent draft with systematic claims and evidence mapping.
- An inventor combines lab notes and code to write a well-grounded technical disclosure for an attorney.
- A patent engineer compares an existing paper against a patent for novelty and consistency audit.
- An R&D team mines patent points from project materials and generates preliminary disclosure documents.
- A researcher needs to document an algorithmic invention with core formulas as editable Office Math in DOCX.
What are this skill's strengths and limitations?
- Enforces evidence grounding, making claims traceable to specific sources.
- Supports multiple input formats (PDF, scanned PDF, text, mixed project).
- Produces complete DOCX patent packages with editable Office formulas.
- Includes validation scripts to catch errors and warnings.
- Skill is in Beta, may have edge cases on real-world inputs.
- Requires Python environment and dependencies, increasing setup bar.
- Generated drafts are not patentability opinions and require professional review.
- Full automation from paper to patent may be slow and require iterations.
How do you install this skill?
Install the full collection via npx or by cloning the repo (e.g., npx skills add ...). This skill requires Python 3 with additional packages (requirements.txt) and Mermaid rendering; for CNIPA search, also install Playwright.
How do you use this skill?
In a compatible agent, describe the paper or materials to convert (e.g., "Turn this paper into a Chinese patent draft"), and the skill will trigger and run the full workflow automatically.