Python patent-infringement analysis skill for agent harnesses (DSH / Claude Code / Codex): input a patent number, get a traceable, reviewer-ready claim-chart-style report in ~1 hour.
DSH integration
Compatible
Author-claimed
Safety audit
Unaudited
Last verified
2026-09-02
License
Not specified by the author
01What can it help you accomplish?
Turn a patent publication number into a competitor-infringement analysis report
A feature-by-feature comparison table, clickable public evidence URLs, and actionable advice on missing evidence and where to search next
IP lawyers and legal teams doing infringement clearance
Compress manual infringement clearance from 3–5 days to about 1 hour
A traceable, reviewer-ready report produced by a 4-module serial pipeline with multi-source search and multimodal evidence capture
Legal-ops and IP teams that need fast, auditable first-pass analysis
02How to install into DeepSeek Harness
Prerequisites
- A host agent that can load skills (e.g. Claude Code, Codex, or a DeepSeek Harness-class agent).
- For standalone mode A: at least one search tool (Tavily/Bocha/Exa/Brave) and a large-model API.
Installation steps
- 01
Recommended (mode B): embed PatentRadar as a skill in your host agent; the host agent spawns 4 subagents and uses its own WebSearch / WebFetch and chat capability — no extra dependencies.
- 02
Alternatively (mode A): run the standalone pipeline driven by server/runner.py, which chains 4 CLI subprocesses.
Verify the integration
- Mode A: open the Web Dashboard and check the 4 module log files to confirm the pipeline finished.
- Mode B: read the analysis report in the host agent's chat window and check the saved JSON on disk.
03DSH integration and capability boundaries
Works with DeepSeek Harness as a host agent: install PatentRadar as a skill (mode B, zero extra dependencies) and let the host agent spawn 4 subagents that call its own WebSearch / WebFetch and chat capability, or run the standalone pipeline (mode A) via server/runner.py and feed results back into dsh.
Multi-source search + multimodal evidence capture
patent number + competitor target→public evidence URLs captured across sources and modalities
Standalone pipeline (mode A)
patent number; needs a search tool + LLM API→report via server/runner.py orchestrating 4 CLI subprocesses
calls external search and LLM APIs, consuming quotaAgent-skill embedding (mode B)
host agent with WebSearch/WebFetch + chat→report produced by 4 spawned subagents inside the host agent
04Who is it for? When not to use it?
Good for
- IP lawyers and legal teams doing infringement clearance
- Legal-ops and IP teams that need fast, auditable first-pass analysis
Not for
- Mode A independently needs at least one search tool (Tavily/Bocha/Exa/Brave) and a large-model API; mode B relies on the host agent's WebSearch / WebFetch and chat capability, so both paths need network access and API quota.
05Compatibility, maintenance and safety notes
- Mode A independently needs at least one search tool (Tavily/Bocha/Exa/Brave) and a large-model API; mode B relies on the host agent's WebSearch / WebFetch and chat capability, so both paths need network access and API quota.
- Reports are LLM-generated for traceable review by lawyers; they are not legal advice and final decisions still require professional review.
Python project tagged dsh-plugin / patent-analysis / legal-tech; 57 stars, 11 forks.
06Frequently asked questions
How do I run PatentRadar inside DeepSeek Harness?
Use mode B — embed PatentRadar as a skill in your host agent (Claude Code / Codex / dsh-class harness). The host agent spawns 4 subagents and calls its own WebSearch / WebFetch and chat capability; PatentRadar needs zero extra dependencies.
What does PatentRadar output from a patent number?
A competitor-infringement analysis report: a feature-by-feature comparison table, clickable public evidence URLs, and actionable advice on what evidence is still missing and where to search next.
Do I need my own search and LLM APIs?
Mode B relies on the host agent's WebSearch / WebFetch and chat capability, so it needs an agent with those tools. Mode A independently needs at least one search tool (Tavily/Bocha/Exa/Brave) plus a large-model API.
How fast is it versus manual clearance?
Manual clearance is typically 3–5 days; PatentRadar compresses the chain to about 1 hour via a 4-module serial pipeline, multi-source search, and multimodal evidence capture.
Can I trust the report as legal advice?
No — the report is LLM-generated and meant to be handed to lawyers for traceable review; final legal decisions still need professional review.
07Related DSH workflows
modlens
by liustack
Vision bridge for text-only models: paste an image, get structured JSON evidence (OCR, layout, semantics).
agent-vision-toolkit
by anionex
Vision toolkit and skills that give text-only LLMs eyes — multi-image understanding, image Q&A, OCR, frontend UI restoration and GUI automation, with optional agent integration.
dsh-vision-router
by ysr666
Eyes for text-only DeepSeek Harness agents: built-in free vision chain (no key) + pixel-level vision tools (Q&A, grounding, crop, pixel diff, colors, OCR, SVG trace, cutout, screenshots). One-command install, no Python, image turns work like ordinary tool-calling turns.
dsh-vision-toolkit
by anionex
为纯文本 DSH Agent 提供 10 个结构化视觉工具：意图感知图片问答、长截图 OCR、原始像素 grounding、UI 还原、像素 diff 等
08Data and sources
零额外依赖，只需要codex或其他agent
由宿主 agent（Claude Code / Codex）的对话能力直接调用
调度 | [`server/runner.py`](src/patentradar/server/runner.py) 串 4 个 CLI subprocess
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-09-02. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.