patent-novelty-evaluator
Skill1 starsChanged 52 days agoMITView on GitHub
Stress-test an invention for patentability (novelty, inventive step, subject-matter eligibility, claim scope) against the closest prior art found. Gives a blunt verdict: dead on arrival, weak, or defensible. Use for "is this patentable?" or to re-check a draft claim. Technical analysis, not legal advice.
In this repository32 other files it gives its agents.
- agent-system-designSkill
- api-contract-designSkill
- architecture-designSkill
- architecture-optimizationSkill
- ci-speedupSkill
- code-handoffSkill
- cross-ai-handoffSkill
- crypto-storage-auditSkill
- database-designSkill
- dataset-sourcingSkill
- distributed-debuggerSkill
- embedded-constrained-codingSkill
- freedom-to-operate-analysisSkill
- frontend-architectureSkill
- github-masterSkill
- idea-finderSkill
- ieee-paper-drafterSkill
- memory-leak-hunterSkill
- ml-model-selectionSkill
- ml-security-auditSkill
- mobile-platform-architectureSkill
- observability-designSkill
- patent-drafterSkill
- performance-optimizationSkill
- project-scaffoldingSkill
- researcher-evaluatorSkill
- researcherSkill
- safety-compliance-scopingSkill
- scale-refactorSkill
- secure-code-reviewSkill
- slow-query-optimizerSkill
- test-architectSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Extract the claim skeleton first,: before searching:
- Distinctive: line 17
If the "new" element is really "known technique X applied to domain Y," say so now.
- Very specific: line 20
Load the references before searching: Read references/patent-search.md and references/legal-tests.md from this skill's base directory in …
- Very specific: line 36
Use references/legal-tests.md for the target jurisdiction.
- Distinctive: line 67
Never round "weak" up to "defensible" to be encouraging.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Patent Novelty Evaluator
Most inventions that feel novel are novel only relative to what the inventor happened to search. This skill finds the prior art that would kill the claim, and says so plainly if it does.
1. Scope and budget
Extract the claim skeleton first, before searching:
- the problem solved;
- the mechanism: how it works, not the goal;
- the single element believed to be new.
If the "new" element is really "known technique X applied to domain Y," say so now. It's the most common failure for software and AI patents.
- Ask for the target jurisdiction (US, Europe, India, or other) if it isn't given, and default to US and Europe. The legal tests differ.
- Search budget: up to 15 searches and 6 full documents read by default; up to 30 and 12 if the user is preparing to file. Stop early once one reference discloses every element of the claim. That settles novelty, and further searching won't change the verdict.
- Load the references before searching. Read
references/patent-search.mdandreferences/legal-tests.mdfrom this skill's base directory in the same step.
2. Search for prior art
Cover all three, using the search guide:
- Patents: by mechanism keywords and by CPC class.
- Non-patent literature: papers, preprints, standards.
- Products: shipped devices, open-source code, datasheets. These are prior art too, and the layer most often missed.
3. Pick the closest reference
Name the one reference, or the smallest combination, that comes closest. A list of twenty loosely related patents isn't an answer.
4. Run the tests
Use references/legal-tests.md for the target jurisdiction.
- Novelty: compare element by element against the closest reference. If one reference discloses every element, the claim is anticipated.
- Inventive step / non-obviousness: would a skilled person combine the closest reference with another known one to get the claim, expecting it to work? Is there an unexpected technical effect?
- Eligibility: is this an abstract idea or algorithm on generic hardware, or a specific technical improvement?
- Claim scope: is it too broad (more prior art in reach) or too narrow (easy to design around)? Name the element to add or remove.
Output
State the verdict first.
## Verdict: Dead on arrival | Weak (needs rework) | Defensible as drafted
<one or two sentences: why>
## Claim skeleton
## Closest prior art
<reference, date, what it discloses>
## Element-by-element comparison
| Claim element | Closest reference | Disclosed? |
## Inventive step
## Eligibility (<jurisdiction>)
## How to strengthen the claim
<the specific narrowing or technical effect that would help, if any>
## Search log
<databases, queries, CPC classes, number of documents reviewed>Never round "weak" up to "defensible" to be encouraging. End with one line: this is a technical assessment, not legal advice, and a patent attorney should review before any filing decision.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

