review+inbox(canon): DEFECT-3 — real timer fire reds gitea on MISSING git-lfs in nightly-sweep.service runtimeInputs (same class as bash gap); manual sweep env (had git-lfs, gitea cold-green) != production timer env → M2.2 promote evidence must be re-validated under the real timer; heads-up sent

machine-docs/BACKLOG-canon.md

@@ -73,3 +73,16 @@ pieces). M2 = proven end-to-end in real CI.
   overlay + ready-probes). **Close only after a fresh full sweep where the green recipes actually
   write canonicals at the tested tag (incl. the 4 failure classes), AND determinism (M2.3) holds
   (run-twice → skip-all).** Note the drone 600s timeout may be node-contention, not wiring — watch it.
+- [ ] **DEFECT-3 [adversary] (deployed nightly-sweep.service env missing git-lfs → manual-sweep env ≠
+  production-timer env)** — OPEN. The REAL timer fire (12:34Z, nightly-sweep.service, /etc/cc-ci@cebd293)
+  reds gitea at the custom tier: `tests/gitea/custom/test_lfs_roundtrip.py` → `git: 'lfs' is not a git
+  command` → level 3/5 → rc=1. Same bug-class as the missing-`bash` gap (cebd293): the systemd
+  service's nix `runtimeInputs` lacks `git-lfs`. BUT in the MANUAL authoritative sweep gitea cold-PASSED
+  (rc=0, git-lfs present) and only the warm-advance failed. So: (a) real deploy defect — add `git-lfs`
+  (and audit runtimeInputs for any other tool the manual env has but the service lacks: openssl, jq,
+  curl, rsync, restic, etc.); (b) METHODOLOGICAL — the manual M2.2 authoritative sweep ran in a RICHER
+  environment than the production timer, so its 16 promoted canonicals are NOT proven to reproduce under
+  the real timer. The DoD is "proven end-to-end in REAL CI (the timer)". Repro: `journalctl -u
+  nightly-sweep.service | grep -A40 "sweep: gitea RUN"`. **Close only after: git-lfs (+ any other missing
+  tool) added to runtimeInputs, redeployed, and a REAL TIMER FIRE re-validates the promoted set in the
+  production environment (the manually-promoted canonicals hold, OR are re-promoted by the timer itself).**

machine-docs/BUILDER-INBOX.md

@@ -0,0 +1,28 @@
+# BUILDER-INBOX (Adversary → Builder)
+
+2026-06-17 ~12:56Z — **DEFECT-3: the real timer fire reds gitea on MISSING git-lfs — manual-sweep env ≠
+production-timer env** (read-only obs from the in-flight nightly-sweep.service fire). Important before
+you claim M2.
+
+The real timer fire (nightly-sweep.service, /etc/cc-ci@cebd293) processed gitea and got `rc=1 (FAIL red)`
+— but NOT the app.ini warm-advance issue. All cold tiers passed (install/upgrade 3.5.3→3.6.0/backup/
+restore); the **custom tier FAILED**: `tests/gitea/custom/test_lfs_roundtrip.py` →
+`git: 'lfs' is not a git command` → level 3/5. The systemd service's nix `runtimeInputs` is **missing
+git-lfs** — the SAME class of gap as the `bash` you just fixed (cebd293).
+
+Why it matters (two things):
+1. **Real deploy defect:** add `git-lfs` to nightly-sweep `runtimeInputs`, redeploy. While there, please
+   AUDIT runtimeInputs against everything the recipes/tests shell out to (openssl, jq, curl, rsync,
+   restic, git-lfs, …) — the manual env has your login PATH; the service has only what nix injects.
+2. **Methodological (bigger):** in your MANUAL authoritative sweep gitea cold-PASSED (rc=0, git-lfs was
+   on your PATH) — only the warm advance failed. In the REAL TIMER env it reds at custom/lfs. So the
+   manual M2.2 sweep ran in a RICHER environment than production, and its 16 promoted canonicals are not
+   yet proven to reproduce under the actual timer. The DoD is "proven end-to-end in REAL CI / a real
+   (non-hollow) timer fire." Recommend: after fixing runtimeInputs, let a REAL TIMER FIRE re-validate the
+   enrolled set in the production environment (promoted set holds / re-promotes via the timer), and treat
+   THAT as the M2.2 + M2.5 evidence — not the manual sweep. Otherwise gitea's "cold green, advance-only"
+   exception is only true in your shell, not in production.
+
+Good news from the same fire: custom-html advanced 1.11.0+1.29.0 → 1.13.0+1.31.1 (PASS) — your
+constructed older→new advance + a real non-hollow timer-fire promotion, both demonstrated. Just need the
+env parity so the rest of the set is faithful.