The data synthesis fixture defines three mock tools with base token costs one hundred twenty, one hundred eighty and three hundred twelve, summing six hundred twelve. At point eight pixels per assigned token the bars measure ninety six, one hundred forty four and two hundred forty nine point six. The names suggest stock data, RSI and a summary but no actual provider or model is connected. Tool payloads and final RSI fifty eight point four are prewritten strings. Exports list the entire configured tool array as tools executed even when execution state is ready with zero steps. That field name is not evidence that any tool actually ran. The query has fifty nine characters. The hypothetical step cost formula adds floor of half that length, or twenty nine tokens, to each tool base. Three such increments add eighty seven, yielding six hundred ninety nine rather than the baked six hundred twelve. A four thousand ninety six context budget minus six hundred ninety nine leaves three thousand three hundred ninety seven. At point zero eight pixels per token the budget is three hundred twenty seven point six eight, used fifty five point nine two and remaining two hundred seventy one point seven six. Character length is not tokenization, and the default proof values do not come from the step formula. No actual context limit or model billing is enforced. The validation function only checks whether each parameter key exists in the argument object. Each default tool declares two keys, so the three tools check six key presences; it does not validate value type, numeric bounds, ticker identity or price sample length. At forty pixels per key the three tool bars are eighty each. Missing Cytoscape aborts launch before auto hydration. Inline reset, presets and actual JSON export remain callable, while Step fails on a null graph before token arithmetic. The exported state is compared directly instead of trusting stale visible proof text. No RSI calculation, market data retrieval, actual tool execution or financial recommendation is claimed.