Patent Claim Analysis
Splits patent claims, charts every limitation against what you build, and names the cheapest design-around.
Free
Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsPatent Claim Analysis
Example session with this skill installed
Analyze US7979296B2. Compare it against a new ticket system that has a GUI and a store but only syncs to one legacy database on a single fixed schedule. Give me a design-around.
- Read your context and instructions
- Compiled the patent claim analysis
Patent US7979296B2: 20 claims total, 2 independent claims.
Element Chart: Claim 1
| # | Claim element (verbatim, shortened) | Accused feature | Status | Note |
|---|---|---|---|---|
| 1 | work item storage holding tasks from a workflow process | ticket store | literal | same function |
| 2 | GUI presenting items and accepting input | dashboard GUI | literal | |
| 3 | interface accepting items from a plurality of independent providers | API connectors | literal | |
| 4 | item type registry tracking handling by type | metadata tags | equivalent | check if it tracks handling logic |
| 5 | custom attribute retriever, staggered sync across two legacy systems with independent timing | single fixed sync | absent | system uses one source and one timing |
Verdict
Claim 1 is not infringed. Element 5 is a conjunctive requirement for "two legacy systems" and "staggered sync," both of which are absent in the accused product. Since Claim 1 is the independent parent, the narrower dependent claims (2-20) are also not infringed by this specific architecture.
Design-Around Recommendations
Maintain Single-Source Sync: The simplest defense is to avoid adding a second legacy source. The claim specifically requires "two legacy systems."
2.
Unified Timing: If a second source is added, use a global synchronization clock rather than "independent timing per connector."
3.
Change Preamble Subject: Move the sync logic to a stateless worker that does not qualify as a "computer readable medium" as defined in the preamble.
Next steps
- Verify if the "metadata tags" in the accused feature truly function as a "type registry."
- Document the single-source architecture in the technical specification as a conscious design choice.
- Note that this analysis lacks prosecution history and does not constitute legal advice.
Connects securely to your tools. The creator never sees your data.
About this skill
Most patent tools summarise the abstract. The abstract does not decide anything. The claims do.
This skill reads a patent the way it has to be read. It fetches the text from Google Patents, separates independent from dependent claims, identifies the transitional phrase that controls scope (comprising is open, consisting of is closed), and turns every limitation into a row you can answer: literal, equivalent, or absent.
What you get back
- Claim count and how many are independent. The fastest read on how narrow the patent really is.
- One element chart per independent claim, with the element, your feature, and a status you can defend.
- A verdict tied to specific rows, not to the abstract.
- A design-around list ordered by cost, from "leave out the missing element" to "licence it".
- For EPO two-part claims, the split between what the applicant conceded was known and what is actually being claimed as new.
It also handles the job people confuse with this one: extracting the architecture a patent teaches, including the stated alternatives in the description that almost nobody reads.
Two scripts, no dependencies
Python standard library only. Nothing to install.
patent_fetch.pymakes one read-only request to patents.google.com and saves the HTML plus plain text.claims_split.pyparses US, EP and WO/OCR claim markup into JSON: number, dependencies, transitional phrase, preamble, limitations.
Tested against granted US patents (US7979296B2, 20 claims), US applications (US20220237532A1, 9 claims), EPO two-part claims (EP1000000A1, 11 claims) and OCR'd WIPO filings (WO2014152996A1, 13 claims).
Why not just prompt it yourself
A generic prompt anchors on the title and the abstract, which carry no legal weight, and it silently skips limitations. The failure mode is confident and wrong. This skill forces the limitations-first order and refuses to conclude before the chart exists.
Known limits
No prosecution history, no jurisdiction-specific doctrine, no legal opinion. It is the technical analysis an attorney wants on the table before the conversation starts, not a replacement for one.
Built by the author of patent-research-mcp, the open-source MCP server for the same pipeline. If you already run it, the skill tells the agent when and how to drive it.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scanned
Verified clean 2 days ago
- Free to download with an account
Requires