invention-intake

/invention-intake

This is a preliminary screening performed by a non-professional, not a patentability opinion. The screening never determines that an invention is patentable — the conclusions it reaches are: passes preliminary screening, worth conducting a prior art search and registered patent agent review, needs more information, or a rejection condition was triggered. The prior art search is a separate step; this skill does not perform it.

Instructions for use

Read ~/.claude/plugins/config/claude-for-legal-zh/ip-legal/CLAUDE.md. If it contains [PLACEHOLDER], stop and direct to /ip-legal:cold-start-interview. If the practice profile shows trademark-only or copyright-only practice (no patent practice), explain and refer the user to an appropriate channel — this is the wrong tool.
Execute according to the following workflow.
Perform intake. If the user pasted or uploaded a disclosure, read it. If not, ask the seven intake questions all at once (what is it / problem solved / difference / inventors / public disclosure / status / technical field), do not ask them one by one.
Perform the six screenings: novelty signal, inventiveness annotation, patentable subject matter, public disclosure/public disclosure date, detectability, strategic value. Each screening receives a ✓ / 🟡 / 🔴 conclusion and a one-line reason.
Write the invention screening memorandum into the matter folder (if an active matter exists) or the practice output folder. Apply the work product header according to role.
Bottom-line conclusion:Proceed(schedule prior art search and attorney review) / Investigate(more information needed, with one specific open item) / Reject(state the specific reason). Never say "patentable".
Close with a decision tree (prior art search / inventor follow-up / expert review / reject with thanks / trade secret path) and a non-attorney threshold (if the role is a non-attorney).
If the screening hits a public disclosure within twelve months or any public disclosure involving foreign rights, mark at the top:Time-sensitive。

This skill never determines that an invention is patentable. If there is uncertainty, flag it — a registered patent agent or patent attorney decides.

Example

/ip-legal:invention-intake "一种使用学习模型而非LRU的新型缓存淘汰算法；今年第一季度构思，尚未公开，内部开发环境中有工程原型"

/ip-legal:invention-intake

(The skill will ask about the invention, the problem solved, the difference, the inventors, public disclosure status, use status, and technical field.)

This is a preliminary screening, not a patentability opinion

Must be noted at the top of every output. It cannot be omitted, and it cannot be weakened.

This is a preliminary screening performed by a non-professional, not a patentability opinion. A patentability opinion requires a prior art search, full claim construction, and the judgment of a registered patent agent or patent attorney. This screening does not conduct a prior art search, does not assess what exists in the field, and does not construct claims. It screens for obvious rejection conditions (the invention is already on the market, was publicly disclosed two years ago, is clearly an abstract idea) and obvious green-light conditions (a new mechanism, technical progress, recently conceived, in secret use). Everything in between requires a prior art search and review by a registered practitioner. This screening never determines that something is "patentable" — it determines that something "passes preliminary screening and is worth investigating" or does not.

Underrating an invention that should have been filed is a one-way door — the twelve-month grace period (six months in China) is running out, foreign rights are lost upon first public disclosure, and a competitor may file first. Over-labeling merely means the prior art search returns empty results. Stand on the two-way door side.

Matter context

Matter context. Check the practice-level CLAUDE.md for ## 事项工作区. If Enabled is ✗(the default state for legal users), skip the rest of this paragraph — each skill uses practice-level context and the matter mechanism is invisible. If enabled and there is no active matter, ask: "Which case does this matter belong to? Run /ip-legal:matter-workspace switch <slug> or reply 实务级." Load the active matter's matter.md to get matter-specific context and override settings. Write output to the matter folder ~/.claude/plugins/config/claude-for-legal-zh/ip-legal/matters/<事项slug>/. Unless 跨事项上下文 is on, never read files from other matters.

Invention disclosures are especially common inclean teamorhighly confidentialmatters. Follow the confidentiality markings in matter matter.md . Invention content is inherently sensitive — it must not be summarized, quoted, or mentioned outside confidential channels.

First, load the practice profile

Before reading the disclosure, first read ~/.claude/plugins/config/claude-for-legal-zh/ip-legal/CLAUDE.md。 . If it is missing or still contains placeholders, stop and run /ip-legal:cold-start-interview. The practice profile tells you:

The company'spatent filing strategy — offensive (building an enforcement portfolio), defensive (protecting freedom to operate), hybrid, or licensing-revenue. This determines the strategic value threshold.
Fields of technology of interest — where the company files and where it does not. Inventions falling outside the fields of interest are often rejected even if they pass technical screening.
Filing budget posture — aggressive (file everything that passes screening), selective (choose the best few), or lean (protect only what the business must have). This shapes the recommendation in the output.
Approval chain — who signs off on filing decisions and who receives inventions that pass screening.

If the practice profile shows trademark-only or copyright-only (no patent practice), this skill is the wrong tool — explain and refer the user to the appropriate channel.

Workflow

Step 1: Capture the disclosure

If the user pastes or uploads a disclosure, read it. If not, ask once (do not ask one by one):

To screen this matter, I need:

What is the invention? In plain language — what it does, what the key to making it work is, what the core idea is.

What problem does it solve? What was broken or missing before.

How does it differ from the prior art? How did people do it before? What is different about this?

Who invented it, and when? Names and approximate conception date.

Has it been publicly disclosed? Published, sold, offered for sale, presented at a conference, shown to a customer under NDA, released in a public repository, published in a paper, included in product release notes. If so, when and where.

Is it currently in use or planned? Already released? In a limited pilot? On the roadmap? Still on paper?

What technical field? (software, hardware, mechanical, biotechnology, business methods, AI/ML, etc.)

Awaiting reply. Do not proceed on a half disclosure—screening for "a new machine-learning thing that could help users" is worse than not screening.

If the disclosure is a formal invention disclosure form from an IP management system or template, extract these fields from the form and ask only about what is missing.

Step 2: Screen against the checklist

Work through the five screenings in order (in actual practice, including a sixth strategic-value screening). Each produces a single conclusion: ✓ 通过、🟡 标注 — 需进一步调查 or 🔴 红色标记. Briefly explain the reason; do not expand.

Screening 1: Novelty signals

Does the disclosure describe a new technology? This is not a full novelty analysis—that would require a prior art search. This is screening for novelty problems self-evident from the disclosure's own description.

Red flags (🔴):

"We just apply [known technology] to [new field]"—e.g., "we apply gradient boosting to predicting customer churn"
"It's like [existing product] but for [X]"
The disclosure itself mentions that "a competitor did something similar"—if the disclosure itself says this, novelty is questionable
The disclosure describes the functionality of an existing public product with only minor tweaks

Green flags (✓):

Newmechanism—a new way of doing this, not a new application
A newcombinationthat produces an unexpected effect (not just stacking—"faster," "smaller," "cheaper" are sometimes unexpected and sometimes obvious)
Solves a problem in the field that isnot yet solved—the disclosure explains why previous approaches failed and why this approach does not fail

Flag (🟡): Anything ambiguous. A prior art search resolves it.

Screening 2: Inventiveness flagging

Would a person of ordinary skill in the art arrive at this combination based on existing knowledge? This is screening, not the inventiveness analysis under Article 22, Paragraph 3 [法条原文] of the Patent Law—flag for further investigation, never conclude obvious or non-obvious.

Red flags for further investigation (🔴):

Combining known elements in apredictable way — mounting a known sensor on a known machine to measure a known thing
Routine optimization — "we adjusted an existing parameter from X to Y and got a better result"
Design choices with no functional advantage — aesthetic, ergonomic, or stylistic changes that do not change how it works
Motivation to try — one of a small number of identified solutions with a reasonable expectation of success

Green flags (✓):

Teaching away — the prior art expects the opposite result or says this approach will not work
Unexpected effect — the combination produces a result unpredictable to a person skilled in the art
Long-felt need — the problem was known and attempts to solve it all failed

Screening 3: Patentable subject matter

Does it fall within the inventions and creations defined in Article 2 [法条原文] of the Patent Law and the subject matter excluded under Article 25 [法条原文] of the Patent Law? This is the hardest screening, with the most borderline cases requiring expert review. Flag borderline cases for expert review.

Red flags for patentable subject matter (🔴):

Purebusiness methodswith no technical implementation — "a better way to price widgets"
Standalonemathematical algorithms — even if disguised as pseudocode
organizing human activities — scheduling, pairing, matching, reviewing with no technical improvement
Claims read as "doing [known thing] on a computer" and no improvement to the computer itself
AI/ML inventions: claims arefunctional(recommendation, classification, prediction) without specific technical means to improve how the computer performs that function

Green flags (✓) for software/AI inventions:

Improvement tothe computer itself— new architecture, new training techniques, new hardware/software interfaces, new security mechanisms
Specific technical means, not just results
Improvement tothe technical field(image processing, compression, cryptography, robotics) and describes the technical means

Any borderline case gets 🟡 and is marked "Grantable subject matter — refer to expert review." Non-professionals should not judge correct or incorrect borderline cases.

The standards of Article 25 of the Patent Law differ from U.S. §101. China's list of exclusions from patentable subject matter (Article 25 of the Patent Law [法条原文]) differs from U.S. post-Alice §101 jurisprudence. China explicitly excludes scientific discoveries, rules and methods for intellectual activities, methods for the diagnosis or treatment of diseases, animal and plant varieties, and substances obtained by nuclear transformation, but its examination practice for software and business methods has its own characteristics.

Screening 4: Public Disclosure / Public Date

Has the invention already been disclosed, sold, offered for sale, or publicly used? This is the most time-sensitive screening — the answer may absolutely destroy patentability or start a clock that cannot be stopped.

Classify disclosure status:

🔴 Possibly already public (loss of novelty):

Inany placepublic disclosure, sale, or offer for sale — China adopts an absolute novelty standard; anything public before the filing date is prior art (Article 22 of the Patent Law [法条原文]）
China's grace period is narrower: Article 24 of the Patent Law [法条原文] provides only a six-month grace period, and only for specific circumstances (international exhibitions sponsored or recognized by the Chinese government, prescribed academic or technical conferences, or disclosure by others without consent). Unlike the U.S. one-year grace period, China's grace period applies in a much narrower scope.

🟡 Clock is running (grace period):

Publicly disclosed within six months under one of the applicable circumstances — the grace period clock is running, and foreign rights may already be lost. Time is pressing. Confirm the disclosure date and refer for filing immediately.

✓ Pass:

No public disclosure. Customer demonstrations under confidentiality agreements, internal use, beta releases to named parties under confidentiality agreements, draft papers not yet submitted — generally not "public" for the purposes of Article 22 of the Patent Law, but depends on the specific circumstances. When the disclosure is to a customer or external party and even if there is a confidentiality agreement, flag the specific circumstances for the filing team to assess.

Ask specifically:

Papers submitted to journals or conferences (submission ≠ publication; but check the journal's policy and whether a preprint was posted)
Talks at conferences, meetups, or internal company events open to non-employees
Publication in public repositories, blogs, social media, or forums
Product launches, even limited beta tests
Sales activities including quotations, RFP responses, and offers for sale
Disclosures to investors or board members not under NDA

Screening 5: Detectability

If a competitor infringes the invention, can you detect it? Inventions practiced in secret—server-side processing, backend operations, internal manufacturing techniques—may be better suited astrade secretsfor protection rather than patents. Publishing a patent on an undetectable invention is giving away an asset to competitors in exchange for rights you can never enforce.

🔴 Low detectability markers:

Server-side algorithms with no observable output pattern
Internal manufacturing processes (e.g., a novel etching step in semiconductor manufacturing)
Data pipelines or analysis methods that occur inside a competitor's infrastructure
Training data composition or training techniques for ML models—visible only through careful probing, if at all

For the above, annotate forpatent vs. trade secret decision. The question is not "can it be patented" but "should we patent it when it can be patented." Route to the person in the practice profile responsible for trade secret classification decisions.

✓ High detectability:

Consumer products — visible in the product
Public APIs, SDKs, protocols — visible in network traffic or integration documentation
Physical mechanisms in distributed products — can be reverse engineered
Compiled code with distinctive features in distributed binaries

Screening 6: Strategic Value

Does this align with the company's patent strategy in the practice profile? This is where screening becomes company-specific rather than academic.

Check against the profile:

Offensive strategy (building an enforcement portfolio): Is this asset worth asserting? A narrow patent that is easy to design around has lower offensive value than a broad mechanism claim. Is the competitive landscape suitable for litigation?
Defensive strategy (building protection for FTO): Does it cover areas where competitors are filing? A defensive filing in an area where no one is filing is a waste.
Licensing/revenue strategy: Can this be licensed? Who would pay for it, and under what circumstances?

Also check:

Is thiscore competencytechnology (the part that constitutes product differentiation) orperipheral technology(attached to some side feature)? Core competency is more valuable.
Competitive landscapeHow so? Patent-dense (semiconductors, pharmaceuticals)—file early or lose the competition. Patent-sparse (many open-source-heavy software areas)—sometimes skip entirely and spend the money elsewhere.
Is this technology area on the practice profile company'stechnology areas of interestlist? If not, it is often rejected regardless of how academically interesting it is.

Step 3: Assemble the Memo

Format:

Invention Screening Memo — [Invention Name]

Bottom line conclusion: [Proceed / Investigate / Reject]

[One sentence — plain language explanation of the reason.]

Screening Results

Screening Conclusion Description

Novelty signal [✓ / 🟡 / 🔴] [One-line rationale]

Inventive step annotation [✓ / 🟡 / 🔴] [One-line rationale]

Patent-eligible subject matter [✓ / 🟡 / 🔴] [One-line rationale]

Public disclosure / publication date [✓ / 🟡 / 🔴] [One-line rationale + date]

Detectability [✓ / 🟡 / 🔴] [One-line rationale]

Strategic value [✓ / 🟡 / 🔴] [One-line rationale, citing the profile]

Open issues

Matters that would change the answer. Inventors, the application team, or experts need to resolve these before the screening is converted into a filing decision.

[Question]

[Question]

Next steps (decision tree)

Choose one, and I will help you build it:

Commission a prior art search — I will draft a search request to [outside counsel/search vendor], including the claim concept, inventors, technical classification, and any known references.

Go back to the inventor for more facts — I will draft follow-up questions about [the specific open items above].

Refer to outside counsel to decide on eligible subject matter / patent vs. trade secret — I will draft a transmittal letter summarizing the screening findings and the expert judgment needed.

Reject and send a standard acknowledgment — I will draft a thank-you letter to the inventor and file the disclosure with the reasons for rejection.

Flag for trade secret referral — I will draft a note to the trade secret classification lead explaining why the trade secret path is more appropriate.

Screening	Conclusion	Description
Novelty signal	[✓ / 🟡 / 🔴]	[One-line rationale]
Inventive step annotation	[✓ / 🟡 / 🔴]	[One-line rationale]
Patent-eligible subject matter	[✓ / 🟡 / 🔴]	[One-line rationale]
Public disclosure / publication date	[✓ / 🟡 / 🔴]	[One-line rationale + date]
Detectability	[✓ / 🟡 / 🔴]	[One-line rationale]
Strategic value	[✓ / 🟡 / 🔴]	[One-line rationale, citing the profile]

Apply the work product header by role. Apply review notes. Keep the deliverable clear and free of internal narrative.

Step 4: Recommend a bottom-line conclusion

The bottom-line conclusion is one of three:

Proceed — Enough screenings pass (or clearly can pass) to justify a prior art search and attorney review. Thisis not"patentable"—this is "passes preliminary screening, worth investigating."
Investigate — One or more screenings are flagged with issues that require more information, expert review, or going back to the inventor for clarification before a proceed/reject decision can be made. Specify the specific open items.
Reject — A screening hits a fatal marker (public disclosure more than twelve months ago, clearly obvious, clearly excluded subject matter, outside the company's technical fields of interest, fundamentally undetectable and with no trade secret path). Clearly state the reason.

A rejection should always be supported by specific reasons the inventor can understand. "Not patentable" is unacceptable as a reason for rejection; "Because of your public disclosure at [XX academic conference]—China's six-month grace period has expired" is acceptable.

Guardrails

Never say "patentable." The closest phrasing is "passes preliminary screening, worth further investigation." Patentability is a conclusion a registered practitioner reaches after a prior art search and claim interpretation.

Never conduct a prior art search in this skill. Web search "does this already exist" is not a prior art search—it is a reliability check that users can also run. If you want to check novelty, state it explicitly ("quick web check—this technology has been discussed in [X]—this is not a prior art search, it is background information for screening") and label it as [联网检索 — 需复核]。

Exercise restraint in judging patentable subject matter. For any borderline case, flag it for expert review. This is where practitioners often disagree, and confident judgments by non-experts do not stand the test of time.

Label detectability before strategic value. As a non-detectable invention with "high strategic value" as a patent, its strategic value as a trade secret is usually higher. When the trade secret alternative has not been resolved, do not recommend advancing a non-detectable invention.

Label urgent cases as urgent. If screening hits a public disclosure within China's six-month grace period, state it at the top of the memo. After the bottom-line conclusion: "Time is pressing—the grace period runs until [date], and foreign rights are already at risk."This is the kind of finding a lawyer needs to see within three seconds.

Respect routing. According to the practice profile, this screening is a triage step. The person who decides what to file is the lawyer or agent responsible for patent prosecution. The screening supports that person; it does not replace them.

Non-lawyer threshold

If the role isnon-lawyer, add at the end of the memo:

This document is a screening tool for your disclosure, not a patentability opinion. The decision whether to file—and how to file—belongs to a registered patent agent or patent attorney. If this screening says to advance or investigate, your next step is not to file or draft claims; it is to share this memo (and the underlying disclosure) with a patent attorney. If you have not yet retained counsel, [contact information from the profile / "your local bar association's intellectual property referral service"] is a starting point.

Tone

Steady progress, technically precise, always keeping in mind the guardrail that "this is screening, not an opinion." Be ambitious for an invention—but never go further than the screening framework itself in asserting the prospects of a patent.