productlisted

# research:product — use the product, then show it working You were asked to research a **product**. The failure mode this skill exists to kill is the report that *read about* the product — a page of accurate prose, one screenshot of an idle landing page, and a feature list nobody watched run. That doc proves nothing. A product is verified by **driving it**: you sign in, you click through the real journeys, you capture each step, you record the headline flow, and you put what the marketing *said* next to what the product *did*. Humans read visually — so the output leads with the product working, not with paragraphs. Product / question: `$ARGUMENTS`. | This skill | Not this skill | |---|---| | **Drive** the product, capture the journey, prove claims visually | `research:research` — answer a text question across engines | | Signs up, clicks through real user flows, screenshots every step, records a clip | `demo` — prove **your own** shipped thing, before/after | | Output: journey diagrams + screenshot strips + a clip + a claims-vs-reality table | `browser` alone — one tab, by hand, no synthesis | ## The two bars you may never skip 1. **Every feature you report is SHOWN being used** — a screenshot of it mid-flow, or a clip. A feature you only read in the docs is a **claim**, labeled as such, never stated as observed. If you couldn't drive it (paywall, version gate), say so — don't launder a doc line into a demonstrated fact. 2. **The journey is captured, not summar