Task: Cover the no-sample families with synthetic data

Every output the seed names is read, from a checked-in sample or from a synthetic fixture that states what its writer writes.

• Every output the seed names is a family with a sample, real or synthetic, or is stated out of scope with a reason, and the generator refuses one that is neither.
• A synthetic fixture's columns come from the writer that produces it at ORE v1.8.17.0, and each fixture names its writer.
• The catalogue marks a family whose only fixture is synthetic, so a reader can tell a sample ORE shipped from one we wrote.
• The corpus gate reads a synthetic fixture exactly as it reads a sample, and fails when a family has neither.

The seed names 44 output files; 27 had no family. The writers at ORE v1.8.17.0 (3d75a6908, the version the project vendors) sort them:

• Real samples outside the suites. runtimes and scenariodump are in the notebook examples, timeaveragedexposure in the legacy ones, and pricingstats is captured from a run here. Each source is read for the named families only, because the notebooks and legacy examples repeat the suites.
• Synthetic. crif, capital_crif, sacva_sensitivity, jacobi and jacobi_inverse have a writer and no sample; each fixture states its writer's columns in order, with two rows, and the catalogue marks it synthetic.
• Other names. A report the run configuration does not name is written under the engine's report name: cashflow.csv is flows and saccrdetail.csv is saccr_detail, which a run here wrote and the catalogue missed. The seed names the initial margin files .txt. A family now carries alias patterns.
• Stems. pca.csv and meanReversion.csv are written per currency, as the pca and mean-reversion families already match.
• No writer. crossgamma, productclass_mapping, creditcurves, parsensi, the aggregation scenario dump and the xvaSensitivity jacobi names are parameters no code reads at v1.8.17.0.
• Not a report. The .dat files are binary data the engine reads back; the .csv.gz cubes are compressed copies of what rawcube and netcube state as text.
• Text-valued. additional_results, marketObjects, riskFactors and the calibrated model's XML have no figure, so no reader here holds them. They are a new task, Store the text-valued outputs.

The generator states each out-of-scope file with its reason and refuses a seed output that is neither a family nor stated, so the list cannot drift.

Every number this task claims about a tree, with the command that produced it and the commit it was measured at. A number without both is not a measurement, because a clean-up record describes a tree and trunk moves under a branch: a consumer count, a gate's component count and a suite's size all change with no code changing, and the reader of a bare number cannot tell which tree it came from. Where the raw output matters, put it in an evidence/ file beside the task and name that file in the Command column.

Manual QA scenarios (scaffolded via compass add test_scenario) that verify this task. Link new ones here as they're created; the scenario doc itself links back via its "Verifies task" field.

Every output the seed names is a family with a sample or is stated out of scope with its reason, and the generator refuses one that is neither. The catalogue gained nine families: four from real samples outside the suites (runtimes, scenariodump, timeaveragedexposure, and pricingstats captured from a run here) and five from fixtures written from their writers at ORE v1.8.17.0, marked synthetic. A family now carries the other names it is written under, which fixed a gap a run here showed: cashflow.csv and saccrdetail.csv were left unread. A run now stores its pricing statistics and run times too. The four text-valued outputs moved to a new task, Store the text-valued outputs.