Implement bounded RF matching candidates from customer one-port Touchstone and component table, Python stdlib only. Input in good.json: Touchstone1.x s1p explicit # Hz|kHz|MHz|GHz S RI|MA|DB R positive-real; strict ascending frequencies, passive |S11|<1, finite numbers; reject2port/v2/noise/unsupported ambiguity. components_csv part,kind(L|C),value_si,tolerance; values positive bounds; tolerance0..0.5; max64components500frequency points; top_n<=10. reference_ohms must equal file reference. Convert complex S11 to Z. Enumerate all two-element L topologies (series then shunt across load, shunt input then series to load), either L or C for each supplied component; no same-part quantity ambiguity: state two copies allowed. Include no-network baseline. Rank by WORST BAND reflection, then nominal mean reflection with deterministic ties. For each candidate evaluate nominal and four tolerance corners; declare tolerance CORNERS evaluated, not mathematically certified worst-case. Produce candidates.csv with parts/topology/value/worst/nominal/baseline; response.csv per sampled frequency for top candidates; network.cir with selected components and honest frequency-dependent measured load representation or one explicit reference-frequency equivalent, never constant resistor for complex frequency load without label; engineering.html tables/plots and report.json. Analytical good100ohm/100MHz load should find near-perfect match at center with50ohm inductive reactance and-100ohm shunt cap. Independent verify use nodal admittance / matrix formulation rather than repeated same impedance formula; check every output spectrum, candidate ranking and required completeness; independent tests analytic matched load, complex load, all three RI MA DB equivalent files, tolerance, malformed truncated input, unknown fields, nonfinite/passivity, missing reference, output tamper. No claim measured board performance, antenna safety, parasitics, component power ratings, continuous-band optimum. No frequency extrapolation. Output deterministic and bounded. CLI and own tests.

Common API: engine.py process(input_path,empty_output_directory)->dict with status PASS; verify(input_path,output_directory)->True or raises ValueError. Write report.json containing product=slug,status=PASS,metrics,summary,limitations plus exact output manifest hashes. No writes outside assigned product folder and temporary directories for tests. Parent fixtures good.json,bad.json,required.json,parent_acceptance.py,SMOKE.sh immutable. No import of mutable parent app code.