legal-skills

Replace billable hours with markdown files and an AI that actually follows them.

This repository is a public, executable playbook for U.S. utility-patent work. Not a chatbot wrapper. Not a “talk to a lawyer” landing page. A set of agent skills — SKILL.md files, checklists, examiner/adversary protocols, intake questionnaires, and deterministic Python — that make a language model do the work a patent boutique currently invoices at several hundred dollars an hour.

The goal is not to “assist” counsel. The goal is to displace legal fees. If an inventor, founder, or in-house engineer can run a 316-item pre-filing audit, a simulated USPTO prosecution loop, and an adversarial design-around attack without opening a matter, that is a win. If a firm that used to charge $18,000 to draft and file a utility application is now competing with a checkout of this repo and a model that costs less than lunch, that is also a win.

Lawyers who are good at judgment, strategy, and courtroom work will still have work. Lawyers whose product is walking a known statute, filling known forms, and billing the walk — this repo is aimed at that product.

---

This is not legal advice

Read this before you use anything in this repository.

This project is not a law firm. It does not create an attorney–client relationship. Nothing here is legal advice, a legal opinion, a patentability opinion, an infringement opinion, a freedom-to-operate opinion, or a representation that any application will be allowed, valid, enforceable, or commercially useful.

• These skills do not file applications, do not sign inventor declarations, do not submit Information Disclosure Statements, do not contact the USPTO, and do not appear before any tribunal.
• Outputs are AI simulations grounded in 35 U.S.C., 37 CFR, the MPEP, current official USPTO texts the agent is instructed to re-fetch, and whatever facts you put in the intake. Models hallucinate cites, miss art, and invent confidence. The scripts exist because models also cannot count, cannot do date arithmetic, and will report “all numerals consistent” after finding 19 of 20.
• Fee amounts, form names, PTO mechanics, and §101 case law move. The audit checklist is a primary-source pre-filing checklist — statutes, rules, MPEP, and current Office practice — not a substitute for the official text. The examination and workaround skills require the agent to refresh statutes, rules, MPEP pages, and examination memoranda on every run. Verify those sources yourself before you rely on a result.
• Some acts remain reserved to a human with the right credentials: signing a declaration under penalty of perjury, making the ultimate duty-of-candor / IDS materiality call, recording an assignment, paying the PTO, and anything that requires a registered practitioner. The skills refuse to do those things without your separate, explicit authorization, and even then they will not pretend they are you.
• You own the consequences. A bad claim set, a missed statutory bar, a false entity-status claim, or an omitted material reference can destroy rights. If the stakes are high enough that you would sue someone over a missed deadline, hire a registered practitioner and use this repo as the thing that makes that person cheaper, faster, and harder to snow.

The GPL-3.0 license already disclaims warranty. This paragraph is the same idea in English: use at your own risk.

That disclaimer is required because the work is real enough to hurt you if you treat it as magic. It is not a retreat from the pitch. The pitch is that most of what you were paying for was process, and process belongs in files.

---

Why markdown, not a firm

Patent work is already a document pipeline:

1. Facts that are not in the draft (disclosure dates, inventorship, entity status).
2. A specification and claims that either enable the invention or do not.
3. A search that either found the dead-ringer or did not.
4. An examiner who will reject under 35 U.S.C. 101 / 102 / 103 / 112 until the claims are allowable or you give up.
5. A competitor who will keep the value proposition and drop the limitation you accidentally treated as essential.

Firms wrap that pipeline in scarcity, letterhead, and a rate card. This repo unwraps it.

The files are the product. If a better model ships next quarter, the same files get cheaper and sharper. If the PTO changes a form, you edit a markdown file instead of waiting for a CLE.

---

What’s here

Three skills. They are designed to be run in order. Each one is a complete agent skill: a SKILL.md the orchestrator reads as its contract, plus the payloads, templates, and scripts that keep the model honest.

legal-skills/
├── patent-audit/         Pre-filing audit. 316 checklist items. Deterministic pre-pass.
├── patent-examine/       Recurrent simulated USPTO examination until two independent examiners clear the same version.
├── patent-workaround/    Adversarial design-around testing. Strengthen only what is supported and patentable.
├── LICENSE               GNU GPL v3
└── README.md             This file

Suggested pipeline

     inventor facts + draft
              │
              ▼
      ┌───────────────┐
      │ patent-audit  │  Is this even fileable? Stop on statutory bars.
      └───────┬───────┘
              │ ranked defects + parsed.json
              ▼
      ┌───────────────┐
      │patent-examine │  Mock prosecution. Amend only after a simulated OA.
      └───────┬───────┘
              │ version-locked "allowable" candidate
              ▼
      ┌──────────────────┐
      │patent-workaround │  Can a competitor keep the value and miss the claims?
      └────────┬─────────┘
               │
               └── iterate examine ↔ workaround until two examiners
                   and two adversaries clear the same packet

patent-examine and patent-workaround will use patent-audit’s parser and mechanical checks if that sibling skill is installed. A prior audit is not a substitute for examination. A prior examination is not a substitute for workaround testing.

---

Install

These are agent skills: a directory with a SKILL.md the model is instructed to follow. Copy the three skill folders into the skills directory of whatever agent you run.

On Windows, from this repo:

Copy-Item -Recurse .\patent-audit, .\patent-examine, .\patent-workaround $HOME\.grok\skills\

Then tell the agent, in plain language:

• Use $patent-audit to audit this draft before I file it.
• Use $patent-examine until two fresh examiners independently find the same version allowable.
• Use $patent-workaround to pressure-test this draft against credible design-arounds.

Python 3 is required for the scripts. The audit parser accepts .txt, .md, .pdf, and .docx. PDF needs PyMuPDF; .docx needs python-docx. Missing optional dependencies fail loudly instead of silently producing a half-parse. The state lockers (cycle_state.py, coverage_state.py) use only the standard library.

Do not point a model at the draft and skip the skill. The value is the protocol, not “an LLM that has heard of patents.”

---

patent-audit — pre-filing audit

When: before you file a utility application, a provisional, or a claim set you are about to spend money on.

What it returns: a ranked defect list with honest coverage accounting — every item is pass, fail, cannot_assess, not_applicable, or not_reached. Silence is never a pass.

What it is not: a rewrite of the application. Audit agents are forbidden to edit the draft. Remediation is a later pass.

The 316-item checklist

The checklist is a primary-source pre-filing audit of a U.S. utility draft — 35 U.S.C., 37 CFR, the MPEP, and current USPTO practice — not a vibe. Every item has an id, part, severity, owner agent, input types, automation class, and a blocking flag. Routing is precomputed in reference/checklist_index.json. The full item text lives in reference/parts/ — those files are the agent payloads. Each item body cites a statute, rule, MPEP section, or current official page.

316 items. 105 are blocking — a fail on one of those stops the audit. 28 are mechanical (code). 169 are assisted (model + structured evidence). 119 are judgment.

26 items can only be answered from intake facts. No amount of document review reaches them (disclosure/bar dates, inventorship knowledge, duty of candor, some foreign and fee facts). If you skip intake, those come back cannot_assess. That is the correct answer. Guessing a first-public-disclosure date is how people lose the U.S. right and then pay a lawyer to confirm they lost it.

Core rules the orchestrator is not allowed to break

1. The draft fits in one context. Every agent gets the whole application. Enablement, antecedent basis, and numeral consistency are whole-document properties. Chunking the spec across agents is a bug.
2. Models do not do arithmetic or exhaustive cross-references. Dates, claim counts, fee tiers, dependency graphs, and numeral reconciliation run in Python.
3. cannot_assess is a valid verdict. Duty of candor cannot be established from any document. An agent that guesses has failed the skill.

Phases

Phase 3 agents

Split E/F/G by the owner field in checklist_index.json, not by file. Workload is uneven on purpose: gate_priority 48, doc_filing 46, doc_claims_law 39, gate_threshold 35, then down to doc_narrowing at 14.

Red team (same batch, not checklist-shaped):

• redteam_examiner — write the first Office Action you would actually issue.
• redteam_design_around — three products that keep the commercial value and miss claim 1. If this is easy, claim 1 is too narrow. No checklist item will tell you that.
• redteam_invalidity — assume it issued. Attack priority, §112, and the art the applicant missed.

Scripts

python patent-audit/scripts/prepass.py --draft DRAFT --drawings DRAWINGS `
    --intake intake.yaml --workdir audit_run

prepass.py refuses to continue when no claims parsed, fewer than four sections were recognized, or the detailed description yielded no reference numerals. That is how you keep eleven agents from confidently auditing a document that isn’t there.

Finding schema

Every audit agent returns only a JSON array to findings/.json:

[{
  "item_id": "G14",
  "verdict": "fail",
  "severity": "Serious",
  "location": "claim 3",
  "evidence": "the said locking member",
  "explanation": "No earlier claim introduces a locking member.",
  "suggested_fix": "Introduce 'a locking member' in claim 1, or recite it in claim 3.",
  "confidence": "high"
}]

evidence must be a verbatim quote from the draft. A finding with no quote is not a finding.

Assets

• assets/intake.yaml — the questionnaire. Leave unknowns null. Null → cannot_assess.
• assets/sample_draft.md — a draft with planted defects, for testing the harness.

Typical wall-clock for a 20–60 page, ~20-claim case is the slowest Phase-3 agent, about 5–10 minutes. Claim sets of 100+ should fan doc_claims_law per claim; keep doc_claims_arch on the whole set.

---

patent-examine — recurrent simulated examination

When: you want the draft stress-tested the way an examiner will actually attack it, then revised until two independent simulated examiners clear the same bytes.

What it returns: a version-locked prosecution record, a final ground-truth application, and a prominent statement that this was a simulation.

What it is not: a USPTO allowance, a validity guarantee, or a one-pass checklist. That last job is patent-audit.

The state model (non-negotiable)

Three artifacts, kept separate:

1. Ground truth — the authoritative editable application. Frozen during an examiner dialogue.
2. Proposed candidate — a complete clean application containing the exact amendments on offer. Arguments alone may leave it byte-identical to ground truth. An examiner may clear only a complete candidate, never a promise to draft later.
3. Prosecution record — Office Actions, responses, interview summaries, reference copies, search logs, issue ledger, versions, hashes.

After a clearance, apply the candidate verbatim to ground truth and verify the content hash. Any unapproved substantive deviation voids the clearance.

python patent-examine/scripts/cycle_state.py init --run-dir examination_run `
  --candidate application=application.docx --candidate drawings=drawings.pdf

python patent-examine/scripts/cycle_state.py snapshot --run-dir examination_run `
  --candidate application=proposal/application.docx `
  --candidate drawings=proposal/drawings.pdf

python patent-examine/scripts/cycle_state.py record --run-dir examination_run `
  --examiner examiner-01 --decision conditional-allowable `
  --document examination_run/examiner-01/notice-of-allowability.md

python patent-examine/scripts/cycle_state.py promote --run-dir examination_run `
  --candidate application=application.docx --candidate drawings=drawings.pdf

python patent-examine/scripts/cycle_state.py status --run-dir examination_run

Phases

Prior-art substance gate

The outer agent does not accept the examiner’s chart on faith. For every 102/103 ground it reads the full claim, the full reference, and reconstructs both arrangements in technical terms. Mapping modes are express, inherent, proposed 103 modification, or POSITA knowledge/evidence. Similar words are not a mapping.

If the mapping looks wrong, the outer agent sends PRIOR_ART_ACCURACY_CHALLENGE before narrowing the claims. The examiner must CONCUR_AND_WITHDRAW, CORRECT_OR_MODIFY, or MAINTAIN_WITH_SUPPORT. Examiner agreement is useful, not dispositive. Protocol failure (ignoring the challenge, silently changing theories) closes that cycle with zero clearance credit.

Do not amend to overcome an unsupported rejection before testing whether the rejection is actually supported. That is how people donate scope.

Examiner decisions

Exact top-line tokens, from references/examiner-protocol.md:

• First action: SIMULATED_NONFINAL_OFFICE_ACTION | ALLOWABLE | INCOMPLETE
• Later: CONTINUED_EXAMINATION | CONDITIONAL_ALLOWABLE | ALLOWABLE | INCOMPLETE | TERMINAL_IMPASSE

INCOMPLETE is the right first move when claims are unreadable, drawings needed to understand the disclosure are missing, or effective-filing-date facts are too thin to know which art applies. Do not ask that examiner to “just search anyway.”

For a pre-filing draft, the planned nonprovisional filing date is used; if none is supplied, the run date is an expressly hypothetical conservative AIA filing date. That assumption enables the simulation. It does not establish a legal priority date.

What ships with the skill

Termination as blocked (not success) when essential evidence is unavailable, the only cure is unsupported new matter, a statutory/priority defect is not curable by drafting, or a required inventor decision is missing. Token pressure is not a success condition.

---

patent-workaround — adversarial design-around testing

When: you want to know whether a competent competitor can keep the product and miss the claims — and whether the gap belongs inside the invention you actually made.

What it returns: a coverage ledger, a claim/disclosure amendment package that survived support and patentability gates, and a list of deliberate gaps you chose to live with.

What it is not: an infringement opinion, an FTO opinion, or proof that “no one can design around this.” Two qualifying adversaries is an exhaustion rule for the workflow, not a market forecast.

Five distinctions the skill will not let you collapse

Do not say a workaround is “covered by the patent” because it shares a theme. Exclusionary scope comes from claims and applicable law.

Classification vocabulary

After independent verification, each candidate is one of:

• not-a-workaround — captured, implausible, or value-destroying
• outside-umbrella — credible avoidance of a different invention
• umbrella-gap-supported — inside the intended contribution and supported enough to consider capture
• umbrella-gap-unsupported — strategically aligned, not possessed/enabled by the disclosure. Blocks a qualifying disposition even if you are willing to tolerate it.
• cannot-assess

Absence from the disclosure does not by itself place an alternative outside the umbrella. That fact controls the support gate, not the strategy gate.

Attack matrix (what the adversary must try)

The competitor subagent is a product architect, not a drafter. It does not propose claim amendments. It does not see the invention-thesis map, the coverage ledger, or the capture strategy. It must generate enough alternatives to cover the viable lenses, then fully develop at least five materially distinct candidates unless fewer are technically credible:

1. omit an element or make it optional
2. substitute structure / material / algorithm / signal
3. change topology, control locus, or component boundary
4. split steps across actors, services, locations, jurisdictions
5. change order, timing, triggering, persistence
6. move work to pre/post-processing, training, manufacture, or the user
7. swap automation ↔ human action
8. move between centralized / distributed / edge / cloud / embedded
9. deliver the outcome through a different statutory class
10. exploit a claimed threshold, range, sequence, or exclusivity
11. combine known public technology around the asserted mechanism
12. test whether another independent claim still captures the result

Decisions: GAPS_FOUND | NO_MATERIAL_GAP | INCOMPLETE.

Integration order (what the outer agent is allowed to change)

Stop at the first defensible remedy:

1. Fix a genuine construction or consistency defect.
2. Revise or add a supported independent claim aimed at the inventive mechanism (including parallel method / system / device / CRM classes when justified).
3. Add supported dependent fallbacks.
4. Improve the summary or detailed description with supported alternatives — narrative is not a claim.
5. Flag a continuation / CIP / reissue / separate filing as a counsel question. Do not assume priority or inventorship.
6. Record a deliberate gap when capture would add new matter, collide with art, or exceed the invention.

For a pending application, every amendment maps to the application as filed. For a pre-filing draft, new technical content comes only from authentic inventor-supplied facts, labeled for inventor confirmation. Never fabricate an embodiment to close a gap. An issued patent is a different posture; do not rewrite it as a draft.

For a pending case, potentially material information discovered in search or attack is routed to a qualified human for IDS/timing triage. The AI does not make the ultimate disclosure decision. Unresolved triage blocks promotion.

State locker

python patent-workaround/scripts/coverage_state.py init --run-dir workaround_run `
  --candidate application=application.docx --candidate drawings=drawings.pdf `
  --candidate intake=intake.yaml --candidate thesis=invention-thesis.md `
  --candidate search=search-record.md --candidate evidence=evidence-bundle.zip `
  --authoritative application --authoritative drawings

Required packet labels: application, intake, thesis, search, evidence. drawings is optional at init and cannot later be silently added or omitted. Qualifying pair is NO_MATERIAL_GAP → no-material-gap, or GAPS_FOUND → only-accepted-gaps. Two distinct fresh adversaries on the same packet hash, then promote. Any material input change resets the count.

scripts/test_coverage_state.py is the unit test for that locker. Run it when you change the script.

What ships with the skill

---

What these skills will not do

• Practice law, form an attorney–client relationship, or carry malpractice insurance for you.
• File, sign, pay, or correspond with the USPTO.
• Invent inventor facts, disclosure dates, or embodiments.
• Treat a clean search as proof of novelty. Unpublished applications exist.
• Treat two simulated allowances as a grant.
• Treat two failed design-arounds as an injunction against the industry.
• Audit a drawing set from text alone.
• Credit a pass for an item nobody touched.
• Let you claim small-entity or micro-entity status because it would be cheaper.

If you need someone to put their registration number on a transmittal, hire that person. Hand them the audit report, the prosecution record, and the coverage ledger. That is still an order of magnitude cheaper than paying them to discover the defects these files already know how to find.

---

Source discipline

The audit checklist is written against 35 U.S.C., 37 CFR, the MPEP, and current USPTO pages, item by item. Those sources still move. The examination and workaround skills require a refresh of:

1. 35 U.S.C. from the House Office of the Law Revision Counsel
2. 37 CFR Part 1 on eCFR
3. Binding Supreme Court and Federal Circuit opinions
4. Current MPEP + later examining-corps memoranda
5. Primary prior-art copies (USPTO Public Search, Patent Center, PATENTSCOPE, Espacenet, dated NPL)

Google Patents is a discovery tool. It is not a cite. Unverified aggregator hits are leads. They are not anticipation.

---

License

GNU General Public License v3.0.

Copy it. Fork it. Improve the checklist when the law moves. Charge money for running it if you want — the GPL allows that. What you may not do is take these files proprietary and sell the walk back to the people this repo exists to stop charging.

If you are a lawyer who thinks this is reckless: good. File an issue with a pinpoint cite and a better item. That is how this gets sharper than any one firm’s form file.

If you are an inventor who was about to spend a year’s runway on a first Office Action: start with patent-audit. Fill the intake. Do not skip the dates.