FlowLeap methods
Fifteen patent work methods (skills) for FlowLeap Patent AI agents. Each method is a folder skills/<name>/ with a SKILL.md and, where needed, a references/ folder. The methods call the FlowLeap patent tools by name (search_patents, get_claims, get_family, examiner_baseline, …). They never run a command-line tool.
| Method | Task |
|---|---|
recipe-prior-art-search |
Prior-art and novelty search; holds the search method the other methods use (references/search-method.md) |
recipe-find-better |
Closer prior art than the examiners cited against a granted patent |
recipe-invalidity-analysis |
Invalidity case with the Fallback Ladder and search rounds |
recipe-freedom-to-operate |
Freedom-to-operate clearance for a product |
recipe-patent-landscape |
Patents of a technology area: players, trends, white space |
recipe-claim-analysis |
Claims broken into elements, keywords and queries |
recipe-claim-drafting |
Claims drafted against the closest prior art |
recipe-infringement-charting |
Element-by-element infringement chart |
recipe-office-action-response |
Analysis of an EP or US office action with response options |
recipe-invention-disclosure |
Invention disclosure form with a novelty check |
recipe-maintenance-fees |
US maintenance-fee status and deadlines |
recipe-patent-to-report |
Dossier of one patent |
recipe-academic-literature-review |
Published research against filed patents |
recipe-audit-report |
Auditable record of AI-assisted patent research |
recipe-custom-dashboard |
Self-contained HTML dashboard from verified patent data |
Install
FlowLeap Patent AI app (firm OS and desktop). The app ships these methods: its image and its desktop build copy skills/ at a fixed commit of this repository. Its office tools are built in; patent_analytics, get_fulltext and the PATSTAT tools need FlowLeap Data.
Claude Code. Add this repository as a marketplace and install the plugin. The methods need the FlowLeap patent tools, for example from the FlowLeap connector (the flowleap plugin of flowleap-ai/flowleap-plugins).
/plugin marketplace add flowleap-ai/flowleap-methods
/plugin install flowleap-methods@flowleap-methods
DeepSeek Harness (dsh). The repository is a dsh bundle (package.json dsh.bundle): dsh/index.js publishes the methods on ctx.skills. It is plain ESM with no build step, so the install needs no build permission. Pin a commit:
dsh plugin add github:flowleap-ai/flowleap-methods#<commit>
Develop
npm ci
npm test
The tests parse every frontmatter as YAML, check each description, and check that no method names a command-line tool, the CLI or a skill outside this repository. They also check that every method carries the two fixed paragraphs of tests/paragraphs.json, that the dsh bundle and the Claude Code manifest publish the same 15 methods, and that the dashboard templates pass their smoke test. CI runs them on every pull request.
License
MIT. See LICENSE.
No comments yet. Be the first to write one.