patent-audit

Audits draft U.S. utility patent applications against a 316-item pre-filing checklist.

391|46|Updated Aug 12, 2026
One-click install
npx skills add https://github.com/gfodor/legal-skills --skill patent-audit-gfodor
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: patent-audit
Source: https://github.com/gfodor/legal-skills/tree/main/patent-audit
Command: npx skills add https://github.com/gfodor/legal-skills --skill patent-audit-gfodor

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pyyaml, pymupdf, python-docx, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Drafting errors, missed statutory bars, wrong inventorship, and blown deadlines can invalidate a patent application before it is ever examined, and catching them traditionally requires an expensive attorney review. This Skill runs a structured pre-filing audit of a draft U.S. utility patent application against 316 checklist items grounded in 35 U.S.C., 37 CFR, the MPEP, and current USPTO practice, returning a ranked defect list with honest coverage accounting. ## Core Features & Use Cases - Deterministic pre-pass: Python scripts parse the draft (.txt, .md, .pdf, .docx) into a structured fact-pack, run mechanical checks (claim dependency graphs, antecedent-basis candidates, reference-numeral reconciliation, abstract length, fee arithmetic), and compute every deadline with weekend/holiday rollover. - Gated multi-agent review: Sequential gates on patentability, inventorship, and priority halt the audit on blocking failures, then parallel agents review the specification, claims, drawings, and filing packet, with adversarial verification of every fatal finding. - Red-team simulation: Examiner, design-around, and invalidity attack agents stress-test the claims beyond the checklist. - Use Case: An inventor with a draft specification and claims runs the intake questionnaire and pre-pass, receives a ranked report of fatal and serious defects (e.g., a claim term with no antecedent basis, a missed grace-period deadline), and fixes them before paying filing fees. ## Quick Start Use the patent-audit skill to audit my draft patent application in draft.pdf against the pre-filing checklist and report any fatal defects.

Frequently Asked Questions about patent-audit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit a patent application before filing?

Run the prepass script with your draft and a completed intake questionnaire, then the skill gates on patentability, inventorship, and deadlines before fanning out review agents over the specification, claims, drawings, and filing packet. The synthesizer produces a ranked defect report with coverage accounting.

What file formats can the patent draft parser accept?

The parser accepts .txt, .md, .pdf, and .docx drafts. PDF input requires PyMuPDF and .docx requires python-docx; missing dependencies produce a clear error rather than silent partial output.

Does this skill file the application or provide legal advice?

No. It does not file applications, sign declarations, submit IDS forms, or contact the USPTO, and it creates no attorney-client relationship. Outputs are AI simulations that must be verified against current official USPTO texts.

Why does the audit need an intake questionnaire?

Twenty-six checklist items, including disclosure dates, inventorship, and duty-of-candor facts, are answerable only from intake facts and not from any document. Without intake, those items are marked cannot_assess rather than guessed.

What happens when a blocking checklist item fails?

The audit stops at the gate and reports the blocker instead of spending effort reviewing a statutorily barred or mis-owned application. Every fatal finding is also adversarially verified by independent verifier agents before it survives into the final report.