diff --git a/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md b/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md index ad1627cf..7eb891d9 100644 --- a/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md +++ b/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md @@ -474,3 +474,8 @@ The lead's one-line answers to the questions H2's develop and review raise (impl builder keeps the unbuildable plans (module, fingerprint, why) in its state `unbuildable` (8 MiB), written when a plan is judged and removed when the module builds, and builder_unbuildable reads that state. Memory is never the truth (P-mesh-owns-the-truth); a tool reads bus state (tools-are-how-a-module-is-spoken-to). P3. +- **H-A-426 (the forge's tokens' vault names, 01:22)** C4. Each token is kept in the vault under its account's name, + `module/gitea/mesh-tools` and `module/gitea/mesh-packages`; the generation is the forge's token name, not the vault's. + P3. Noted for after H2 (ยง819's check): the integration count takes every test the integration run reports, untagged + ones and subtests included, so a fold hides rows behind a t.Fatal; the check should count the tagged top-level test + functions, and a module's fold then needs no plain loop.