patent-novelty-evaluatorlisted

# Patent and Novelty Evaluator Most inventions that feel novel to their inventor are novel only relative to what the inventor happened to search. This skill's job is to find the prior art that would actually kill the claim, and say so plainly if it does. ## Step 1 — Extract the claim skeleton before searching anything Force the invention into claim form first, even roughly: - The problem being solved. - The structure or method (the actual novel mechanism — not the goal, the *how*). - The single element the applicant believes is the distinguishing contribution. If the "distinguishing contribution" is vague or is really just "applying known technique X to new domain Y," flag that immediately — it's the most common software/AI patent failure mode and worth naming before spending a search budget on it. ## Step 2 — Prior art search (see the `researcher` skill for search methodology) Cover, at minimum: - **Patent literature**: Google Patents, Espacenet, WIPO PatentScope, USPTO full-text — search by mechanism/function, not just the applicant's own vocabulary. This is non-negotiable; skipping patent search is not a novelty search. - **Non-patent literature**: papers, preprints, standards docs describing the same mechanism. - **Commercial prior art**: shipped products, open-source repos, datasheets doing the same thing in practice, even without a patent behind them — these count as prior art too and are frequently missed. ## Step 3 — Pick the closest reference(s), not a pile of