These cases check that the skill activates on ledger-hygiene work (chart of accounts, categorizing a bank feed, double-entry posting, reconciliation, cash-vs-accrual) and stays out of the way for invoicing, pricing, forecasting, finance interpretation, and Stripe API work.
\nHow to run: there is no automated runner here — cases.yaml is read by the catalog's shared eval harness, or reviewed by hand. For each should_trigger prompt, confirm this skill would fire; for each should_not_trigger, confirm the request routes to the named sibling instead. For the capability case, run the scenario against the skill and grade the response against every line in must_include — all must be satisfied (correct five-category COA, debits=credits on every entry, owner draw as equity, loan principal as liability reduction, customer prepayment as unearned-revenue liability, and a reconcile-to-bank reminder). This is a process/judgment skill, so the capability rubric is the rigor mechanism — there is no verify.sh.