Finish CUST-WP-0072: coverage evidence, templates, intake join.

Record the flavor backfill cohort, residual-intake join, and agent
convention that default views omit residuals.

docs/evidence/2026-09-14-flavor-backfill/README.md

@@ -0,0 +1,67 @@
+# CUST-WP-0072 flavor and depends_on backfill
+
+Date: 2026-09-14. Gate: STATE-WP-0092 is `finished` in files and on the hub
+(`2a2d8bcd`). Live primary schema is still `b5e6f7a8c9d0` (flavor column not
+applied), so default views do not yet drop residuals. Founder instructed
+file backfill to proceed. Hub views follow after alembic `c6f7a8b9d0e1` is
+deployed.
+
+## Classification rules (no auto-promote)
+
+| Flavor | When |
+| --- | --- |
+| `residual` | `origin: residual\|handoff`, or title/body already names a residual handoff |
+| `planning` | already set, or `status: proposed` without implement/deploy verbs, or planning keywords |
+| `refactoring` | title keyword refactor/rename/terminology/cleanup/strangler |
+| `extension` | title keyword adapt/extension/follow-on (and not implement) |
+| `implementation` | default for ready/active/blocked functional work |
+
+Unset was treated as relevant, then filled. Existing valid `flavor:` was kept
+(`CUST-WP-0072`, `INFD-WP-0002`, `COORDINATION-WP-0005`, `RAIL-FAB-WP-0030`).
+`depends_on` was copied from existing `depends_on` / `depends_on_workplans` /
+`blocked_on` workplan ids only — no edges invented from `related:`.
+
+WHITEHAT-WP-0008 was corrected from `implementation` to `residual` (title
+“live ASM residuals after WHITEHAT-WP-0007”).
+
+## Open-workplan cohort (116)
+
+| Flavor | Count |
+| --- | ---: |
+| implementation | 81 |
+| planning | 21 |
+| residual | 11 |
+| extension | 2 |
+| refactoring | 1 |
+
+Residual open workplans:
+
+| Repo | Id | origin_ref |
+| --- | --- | --- |
+| railiance-fabric | RAIL-FAB-WP-0028 | STATE-WP-0079 |
+| railiance-infra | RAIL-HO-WP-0013 | RCLK-WP-0005 |
+| ops-warden | WARDEN-WP-0039 | HFACT-WP-0001 |
+| approval-engine | APPROVAL-WP-0002 | APPROVAL-WP-0001 |
+| tenant-engine | TEN-WP-0012 | TEN-WP-0011 |
+| reuse-surface | REUSE-WP-0021 | IDENTITY-WP-0004 |
+| hub-core | HUB-WP-0011 | HUB-WP-0010 |
+| hub-core | HUB-WP-0009 | OPS-WP-0003 |
+| whitehat-security | WHITEHAT-WP-0006 | WHITEHAT-WP-0001 |
+| whitehat-security | WHITEHAT-WP-0008 | WHITEHAT-WP-0007 |
+| fin-hub | FIN-WP-0006 | RESOURCE-WP-0005 |
+
+Machine table: `coverage.csv` / `coverage.json`.
+
+## T04 — residual intakes vs workplans
+
+42 intakes with `origin: residual`. Join by `origin_ref`:
+
+- 38 → finished/archived parent workplan (intake is the live leftover; not a
+  second implementation backlog)
+- 4 → non-workplan refs (task/decision/legacy id); not open workplans
+
+Zero open workplans were both `flavor: implementation` and the target of a
+residual intake. Provenance stays on the intake; flavor is the live
+classification on the workplan.
+
+See `intake-join.json`.