The corpus in sources/jq.test is byte-for-byte upstream and is never edited. Where a case cannot run under this kit's harness, it is named here instead, so every skip is visible, reasoned, and auditable against the upstream file. This file is a scoring asset: it is hash-verified against the import and restored before grading. One verbatim program line per entry. An entry that matches no case in the corpus is a hard error (exit 2), not a shrug — a silent no-op would quietly re-admit a case the kit cannot run if the corpus pin ever moved. Only the loader cases are excluded. The module *grammar* cases — `module (.+1); 0`, `module []; 0`, `include "a" (.+1); 0`, `include "a" []; 0`, `include "\ "; 0`, `include "\(a)"; 0`, and `%::wat` — remain in the scored set: they are parse errors that a correct front end rejects without ever touching the filesystem.