Market Patent Diligence

Quick start

1. Require either a bounded research_asset or a specific technology_thesis plus target market or diligence question. If both are missing, ask for the paper, abstract, invention summary, product idea, patent/application number, assignee, or market hypothesis.
2. Classify the request before searching: quick scan, patent landscape, market and competitor scan, regulatory/procurement/funding scan, full diligence pack, or red-team evidence audit.
3. Create a search plan from the claim: technical nouns, synonyms, adjacent terms, use cases, customer workflow terms, incumbent categories, IPC/CPC/NAICS/SIC/MeSH terms where useful, jurisdictions, date range, and source classes.
4. Use current public sources for changing facts. Prefer primary or official sources over summaries: patent-office databases, assignment/register data, company pages, public filings, procurement notices, regulator/reimbursement pages, grants/funding databases, clinical or standards registries, and customer-visible pricing.
5. Log every nontrivial source and query. Do not present market size, competitor status, patent ownership, legal status, regulatory status, procurement demand, reimbursement, funding, or pricing claims as facts without a source and retrieval date.
6. Separate source-backed fact, user-provided fact, inference, hypothesis, and speculation in the output.
7. For persistent work, initialize a diligence pack with python3 scripts/init_diligence_pack.py <case-name> --path <output-dir> and validate it with python3 scripts/validate_diligence_pack.py <case-dir> before returning artifacts.
8. End with a handoff: what the evidence says, what it does not say, the weakest commercialization assumptions, and which sibling skill should consume the results next.

Modes

Quick scan

Use for a bounded 30-90 minute-style pass. Produce a concise source-backed read on market/competitor/patent signals, strongest evidence, red flags, and next searches. Do not imply exhaustive coverage.