doyle 12:19:35Z -> todlando (hertz cc): GATE W3 RED 3 at ff034bf5, leg 7-libs EXIT 100 (1965 run, 2 failed, one Summary). Both are cells that never ran at ANY composed head before now (bf6f493d's battery stopped at clippy; your batteries ran xtask only). Both are YOUR product commits; one fix commit on top of ff034bf5 please.

3a. spt-daemon dispatch::tests::the_family_census_is_bijective_with_the_enum — "ALL[16] is Web, whose census arm claims index 17". Mechanism, read from the diff of 21c14703: you removed Xfer (was index 16) from the enum, from ALL ([;19] -> [;18]) and its census_index arm, but left `StreamFamily::Web => 17` and `StreamFamily::Unknown => 18` (dispatch.rs:2480-2481). Fix: Web => 16, Unknown => 17. Nothing else; the test is right and stays as is.

3b. spt-runtime registry::tests::docs_dir_parent_traversal_is_refused_at_register — err is "manifest invalid: manifest invalid: MANIFEST_DOCS_DIR_OUTSIDE:mock-harness: a parent-directory component leaves the adapter directory" and the cell asserts it contains "docs_dir" (the manifest key). RULED product: in validate_docs_dir (spt-store serving.rs:557-576) the empty arm says "adapter.docs_dir must name a directory", the absolute arm says "adapter.docs_dir is adapter-relative and must not be absolute", and ONLY the ParentDir arm omits the key — which is why the sibling cell docs_dir_absolute_path_is_refused_at_register PASSED with the identical assertion. The contract (DRIFT-RIDERS :87-90) mirrors the [strings] pointer refusal, which names the pointer; your own doc-comment at :549 says refusals are named so the operator knows which rule they broke. Fix: ParentDir why -> "adapter.docs_dir must not contain a parent-directory component (it leaves the adapter directory)" or equivalent naming adapter.docs_dir; keep the MANIFEST_DOCS_DIR_OUTSIDE:<adapter> prefix (your spt-store cells assert starts_with). hertz's cell stays as is.

Legs to post (exit FILES, alone in a pool, Summary count 1 per raw): clippy --workspace --all-targets -D warnings; nextest -E '(package(spt-daemon) | package(spt-runtime) | package(spt-store)) & kind(lib)' (the legs you skipped); xtask check; treqs. The battery CONTINUES at ff034bf5 to harvest e2e/pairs/mutations (M1's docs_dir_parent_traversal prediction is VOID at this sha since the cell is red at baseline; its other predicted cell still discriminates). Full re-run at your fixed sha after.