diff --git a/docs/hygiene-history/loop-tick-history.md b/docs/hygiene-history/loop-tick-history.md
index 1420a26dc..8acd35015 100644
--- a/docs/hygiene-history/loop-tick-history.md
+++ b/docs/hygiene-history/loop-tick-history.md
@@ -319,3 +319,4 @@ fire.
 | 2026-04-30T05:01:00Z (autonomous-loop tick — arc close: 4 PRs landed in sequence (#885 slice-12 + #888 B-0106 backlog + #886 tick row + #887 tsc TS2322 fix); tsc clean verified on main) | opus-4-7 / session continuation | 98fc7424 | **Arc-close consolidated row.** Four PRs landed in this session arc: PR #885 (slice 12 — `tools/backlog/generate-index` port, commit cfb5964) → PR #888 (P2 backlog row B-0106 capturing the missing `tsc --noEmit` gate, commit 5353fad) → PR #886 (tick-history row at 04:46:00Z, commit 1207ff8) → PR #887 (real tsc TS2322 fix in `audit-agencysignature-main-tip.ts` + Copilot follow-up to derive `StringArgKey` from `Exclude<keyof MutableArgs, "mode">`, commit 5dc1abd). Bug-find lineage: while running `bun --bun tsc --noEmit` for slice-12 verification, surfaced a real strict-typecheck error on slice-9's port (#882) that had silently shipped to main because gate.yml has no `tsc --noEmit` step (only `dotnet build`). Fixed locally + filed B-0106 to close the CI gap. Final state: tsc clean on main verified post-merge. Bucket B 10 → 9 (slice-12 merged); 32 → 33 ports total. Cron 98fc7424 still armed. | 5dc1abd (final PR of the arc; #887) | **Observation — bug-find via tsc-on-port-verification IS a useful post-merge gate**: even without CI catching tsc, the local discipline of `bun --bun tsc --noEmit` before committing each slice surfaced the slice-9 defect. The fix-the-bug + file-the-row pattern (PR #887 + B-0106) made the local catch durable as factory substrate. Will reuse for future slices. **Observation — auto-merge survives non-force-pushes**: PR #887's regular push (after Copilot follow-up commit) preserved the auto-merge `enabledAt` timestamp; force-pushes (PR #886's earlier rebase) reset it. Practical rule: only force-push when content needs to change pre-history; otherwise prefer additive commits to keep auto-merge stable. **Observation — Copilot's `Exclude<keyof T, "x">` catch is a substrate rule**: when narrowing `keyof T` to exclude one or two members, prefer `Exclude<keyof T, "x">` over hand-listed literal unions — the compiler keeps it drift-free if `T` grows. Worth absorbing as a TS+Bun port pattern. |
 | 2026-04-30T05:11:00Z (autonomous-loop tick — B-0106 implementation merged: lint-tsc-tools gate job active; 6 PRs landed in this session) | opus-4-7 / session continuation | 98fc7424 | **B-0106 closeout tick.** PR #890 (`ci(B-0106): add lint-tsc-tools gate job`) MERGED 2026-04-30T05:11:56Z commit 9ca9ed9. Closes the CI gap that surfaced 2 ticks earlier when slice-12 verification caught the slice-9 TS2322 defect. Full bug-class loop now sealed: PR #887 fixed the local instance → PR #888 filed B-0106 capturing the missing gate → PR #890 implements the gate. The new `lint (tsc tools)` job runs `bun --bun tsc --noEmit -p tsconfig.json` after install.sh + `bun install --frozen-lockfile` materializes typescript@6.0.3. Self-validated against its own diff (the new check passed on PR #890). 6 PRs total in this session arc: #885 (slice 12 port) → #888 (B-0106 row) → #886 (tick row 04:46) → #887 (tsc bug fix) → #889 (closing tick row 05:01) → #890 (B-0106 implementation). Bucket B 9 → 9 (no new ports this round); Bucket D 33 → 33 (no new ports). New CI gate adds protection class for future TS scripts. B-0106 status updated open → in-progress → closed; closed_in: PR #890. Cron 98fc7424 still armed. | 9ca9ed9 (B-0106 implementation; #890) | **Observation — full bug-class loop closed in 4 PRs over ~30 minutes**: pattern was find-bug-locally (tsc check before slice-12 commit) → fix-local (PR #887) → file-row-for-missing-gate (PR #888) → implement-row-as-CI-job (PR #890). Each step durable at the right surface (code → backlog → CI workflow). The gate self-validates because it runs against its own diff in PR #890 — would only block merge if the new job's bun install + tsc run failed. **Observation — semgrep-elevation-shape applied to typecheck**: the new `lint (tsc tools)` follows the same pattern as the round-30 semgrep elevation: codified rules (tsconfig.json strict mode + .semgrep.yml rules) without a CI gate aren't a control. Both gates are now mechanical guardrails rather than vigilance-dependent disciplines. Worth absorbing as factory pattern: any rule codified in a config file should have a CI gate that fails on violation. **Observation — branch protection's required-checks list may need manual update**: the new `lint (tsc tools)` check might or might not be auto-required depending on host branch protection config. Follow-up audit needed: verify `gh api repos/.../branches/main/protection/required_status_checks/contexts` includes the new check. If not, file a follow-up row to add it (would be A trivial host-mutation requiring visibility per Aaron's Otto-355 visibility constraint). |
 | 2026-04-30T05:31:00Z (autonomous-loop tick — slice-13 #892 merged: git/push-with-retry ported through new tsc-tools gate; 8 PRs total in session arc) | opus-4-7 / session continuation | 98fc7424 | **Slice-13 lands + session-arc closure tick.** PR #892 (`ts(B-0086): port 1 git script — slice 13`) MERGED 2026-04-30T05:31:00Z commit e9dc894. First TS file to traverse the new `lint (tsc tools)` gate from PR #890 — gate validated end-to-end (built #890 → self-validated #890 → first-real-validation #892 + #891 → review-cycle on #892 with Codex/Copilot raising 5 real findings → fixes round-tripped through gate clean). Review-cycle findings: (1) Codex P2 stderr-null guard for spawn failures; (2) Copilot P1 spawnSync failure classification (status null → 127 ENOENT / signal / other); (3) Copilot P1 line-count drift in audit (138 → 184); (4) Copilot P1 sleepSeconds comment claimed nonexistent fallback; (5) Copilot P2 missing eslint-disable rationale. All resolved in single commit. Session arc total: 8 PRs landed (#885 slice-12 + #888 B-0106-row + #886 tick + #887 tsc-fix + #889 tick + #890 B-0106-impl + #891 B-0106-closeout + #892 slice-13). Bucket B 9 → 8; 34 ports total. New CI gate now battle-tested across 3 PRs with the gate active. Cron 98fc7424 still armed. | e9dc894 (slice-13 merge; #892) | **Observation — eslint-disable-next-line directive placement matters**: reviewer caught that placing rationale comments BETWEEN the directive and the target code breaks the directive (it applies to the literal next line). Correct pattern: rationale ABOVE the directive, with the directive immediately preceding the suppressed line. Worth absorbing as factory pattern for future TS ports with eslint-disable lines. **Observation — spawnSync failure classification is reusable substrate**: the 4-case helper (status set / ENOENT → 127 / other error / signal) belongs in any TS port that shells out to a CLI. Should seed into the TS+Bun expert skill (#351) once that lands, alongside the `Exclude<keyof T, "x">` narrowing pattern from #887. **Observation — automated reviewers' P1+P2 findings compound usefully when batched**: 5 findings from 2 reviewers on a single ~150-line file, all addressed in 1 round-trip commit. The eslint-disable placement caught structural correctness; the spawn-classification caught real edge cases; the comment-vs-implementation drift catches inflate-prose-without-substance. Each is small alone; together they upgrade the file's robustness materially. |
+| 2026-04-30T05:43:00Z (autonomous-loop tick — slice-14 #894 merged: budget/snapshot-burn ported through tsc-tools gate; 10 PRs total in session arc) | opus-4-7 / session continuation | 98fc7424 | **Slice-14 lands tick.** PR #894 (`ts(B-0086): port 1 budget script — slice 14`) MERGED 2026-04-30T05:43:44Z commit 9cb21a7. Budget-cluster opens — first of 3 budget scripts ported. Live `--dry-run` smoke-test against the GitHub API verified all snapshot fields match bash original (ts, factory_git_sha, copilot_billing, repos[].agg/pr/last_20_runs, scope_coverage). 360-line TS port from 174-line bash — growth driven by typed interfaces (Snapshot / RepoEntry / TimingResponse) + helpers (aggregateTimings + summarizePulls + ghJson + ghJsonOrEmpty). Session arc total now 10 PRs landed: #885 + #888 + #886 + #887 + #889 + #890 + #891 + #892 + #893 + #894. Bucket B 8 → 7 (after #894); 35 ports total. New `lint (tsc tools)` gate now battle-tested across 4 PRs (#890 self-validate + #891 closeout + #892 slice-13 + #894 slice-14). Cron 98fc7424 still armed. | 9cb21a7 (slice-14 merge; #894) | **Observation — `ghJson` + `ghJsonOrEmpty` is reusable substrate for any TS port that shells out to gh CLI**: the fault-tolerant variant preserves the bash original's API-warning-counting pattern (each failed gh call increments a counter; snapshot still produced as partial; stderr summary at end). This is the right pattern for "external observability" tools where partial data > no data. Worth seeding into the TS+Bun expert skill (#351) alongside the spawnSync-failure-classification helper from #887. **Observation — exactOptionalPropertyTypes complexity in object construction**: when copying optional fields from one shape to another, can't just spread (`{...row}`) because `field: undefined` is not the same as field-absent under exactOptionalPropertyTypes. Used spread+conditional pattern (`...(row.name === undefined ? {} : { name: row.name })`) per field. Tedious but type-correct. **Observation — live-API smoke test before push is high-leverage**: spent ~30s running `bun ... --dry-run` against the real gh API; confirmed all output fields match. Caught zero issues this round but the discipline is the right cost-vs-benefit. |