AGENT SKILL

prior-art-scan

Search patents, papers and preprints together to see whether an idea is already known. Use when a team asks if their invention is novel, before filing, before committing engineering effort to an approach, or when preparing to meet a patent attorney.

License 许可
MIT
所属仓库
LNdme/SUniv
最近更新
2026-11-12
Stars
★ 1
Forks
0

01 前言 Front Matter

name
prior-art-scan
description
Search patents, papers and preprints together to see whether an idea is already known. Use when a team asks if their invention is novel, before filing, before committing engineering effort to an approach, or when preparing to meet a patent attorney.

03 SKILL.md 正文

prior-art-scan — what is already known, from every direction at once

An examiner does not care whether prior art appeared in a patent, a journal, a preprint or a conference proceeding. A team searching only one of those is building a false picture of its own novelty.

Search all of it, and say so

suniv-patent "<technical terms>" --limit 25
suniv-search "<technical terms>" --limit 25
suniv-search "<technical terms>" --from <year> --sort year --limit 15

Patent language is not paper language. The same invention is a "graphene-based gas sensor" in a paper and "a sensing element comprising a carbon monolayer" in a patent — deliberately broad, because breadth is what a claim buys. Search both vocabularies: run the team's own words first, then re-search with the phrasing that comes back from the patent hits.

Search the function, not just the implementation. Prior art that solves the same problem a different way still bears on obviousness.

What the results can and cannot tell you

suniv-patent returns bibliographic data. A patent's scope is defined by its claims, and the claims are not retrieved. An abstract describes what the invention is about; it does not delimit what is covered. Never tell a team they are clear of a patent, or blocked by one, from an abstract. Point them at the publication number and say the claims need reading.

Two silences that are not evidence of absence:

  • The eighteen-month window. Applications publish about a year and a half after filing. The most relevant prior art may exist and be invisible today. This alone makes any agent-run search provisional.
  • Vocabulary. Not finding something usually means not having guessed the right term.

So report the outcome as a search, with its bounds: what was queried, in which databases, with which terms, and what came back. "I found nothing" is a statement about the search. "There is nothing" is a claim no one can make here.

Reading what you find

For each relevant hit, say plainly why it bears on the idea:

  • the same thing — same problem, same mechanism;
  • the same mechanism elsewhere — which is what makes an invention obvious, and is the most commonly missed category;
  • the same problem, differently solved — context for the argument, and often the strongest position the team can build from;
  • adjacent — worth knowing, not a threat.

Distinguish what is granted and in force from what is merely published or abandoned; an expired patent is free to use, and teams routinely mistake one for the other.

Where a patent and a paper describe the same work, say so — an inventor publishing their own work is common, and it collapses two apparent hits into one.

Close by handing the work forward

Give the team what their attorney will ask for: the closest art with publication numbers, the terms searched, the databases covered, and where the picture is thin. Then say clearly what this was not — not a freedom-to-operate opinion, not a clearance search, not legal advice. It is the preparation that makes the paid hour productive.

If the idea looks novel and the team is about to publish, disclosure-guard applies immediately, and it is more urgent than anything in this skill.

04 同仓库更多 Skills

LNdme
LNdme/SUniv

use-shared-credential

When you need to call a service the org has a SHARED credential for (a SERP/search key, a paid-API key, a data-vendor feed, a Git remote) — and the platform has told you that credential is available to this conversation — make the call BY PROXY through the credential broker. You never see the secret; the core stamps it onto the outbound request for you.

★ 1
LNdme
LNdme/SUniv

taste-skill

Design taste and process for anything a person will look at in a browser — landing page, dashboard, prototype, deck. The default design driver for a browsable artifact, with an anti-slop playbook for the visual language.

★ 1
LNdme
LNdme/SUniv

publish

Publish a long-lived internal web app, site, or dashboard from the agent computer. Scope-bound (only the owner's scope, plus whoever you share it with, can reach it), immutable versions with rollback, a stable friendly link. No public URL.

★ 1
LNdme
LNdme/SUniv

popular-web-designs

54 real-world design systems (Stripe, Linear, Vercel, Notion, Apple…) as ready-to-paste HTML/CSS — exact colors, typography, components, and spacing for styling a page after a known brand.

★ 1
LNdme
LNdme/SUniv

morning-digest

Assemble and deliver one short morning digest for the person who owns the cron — what changed overnight in the sources they already connected, and what it means for their day.

★ 1
LNdme
LNdme/SUniv

memory

Deliberately search, add to, or curate your long-term memory with the `memory` tool — beyond the automatic recall/capture every turn already does. Use when asked who someone is or what they're working on, when you learn a durable fact worth keeping now, or when your notebook holds stale or wrong entries you should clean up.

★ 1