## A higher Jacobi-sum recurrence must handle a trivial prefix product What happened: this Fermat-polynomial construction requires sums of three or five nontrivial characters whose total product is nontrivial, but an intermediate prefix product can be trivial. Applying only the nontrivial-product binary recurrence would change exact coefficients. What the skill says now: verify substantive identities independently and investigate normalization; it does not prescribe character-sum recurrences. What it should say: a task-specific computation reference could remind contributors to check the hypotheses of intermediate reductions as well as the final expression. For higher Jacobi sums with every original character nontrivial, if the current prefix product is trivial, its sum-at-zero is minus (Q-1) times its sum-at-one. Appending another nontrivial character therefore multiplies the sum-at-one by -Q. Otherwise multiplication by the ordinary binary Jacobi sum applies. This is a proposed example, not an instruction to expand the live core with a family-specific algorithm. Evidence: T729 generate.py has the explicit s==0 branch. independent_check.py compares all 48 complete polynomials with an additive Gauss quotient, using different character evaluation and a different finite-field generator. check-results.json records zero failures, positive known-answer controls and deliberately wrong inputs rejected. No skill source was edited.