[CLAUDE] Memory: sleep-compress L2 test-specialist — activity-s51-s52.gist.md additive

/sleep-recovery-memory-l2 all (S88). Phase 0 scan: 18/19 L2 verbatim SKIP
(14x 2026-05-qN.md already in aggregate 2026-05.gist.md gen:1, verified by
reading source-decl; 4x 2026-06.md double-distill guard gen 1-2; 5 subs
empty archive). 1 distilled: test-specialist activity-s51-s52.md (gen-0
orphan, no gist) -> activity-s51-s52.gist.md (distill-gen:1), 4 records ->
3 clusters (C-A merges the 2 gotcha#57 episodes).

reviewer Fidelity-gate FAIL v1 -> re-distill: C-A merge had mis-attributed
the whole fix to "Mig 45"; verbatim ties Mig 45 to HRM-only (LeaveTypes/
ShiftPatterns), Master (Dept/Project/Supplier) = Mig 47 per canonical
CLAUDE.md. Split per-episode -> PASS. coverage-diff 26/26, G-009 clean,
ratio 97% (signal-dense, report-not-target). last_sleep_at -> 2026-06-25.

+reviewer self-harvest: merge-distill false-specificity anti-pattern
(presence-gate blind to cross-episode shared-token mis-attribution).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

.claude/agent-memory/memory-budget.json

@@ -1,7 +1,7 @@
 {
   "_note": "Harness-9 (S70, 2026-06-17) memory budget. Caps SEEDED BY MEASUREMENT (scripts/measure-agent-memory.ps1), NOT imagined headroom. Budget-audit @session-start (session-start.md §2.1.2): if curate-to-fit is dropping important markers, BUMP the relevant cap rather than cut markers. Re-measure with the script; never hand-edit measured_bytes.",
   "seeded_date": "2026-06-18 (S71 re-measure after G1 curate — reviewer 36.7KB + investigator 29.8KB over-cap from S71 same-role race, curated L1->L2 back under auto-inject cap)",
-  "last_sleep_at": "2026-06-18",
+  "last_sleep_at": "2026-06-25",
   "_last_sleep_at_note": "Harness-10b sleep-recovery (S72): timestamp lan cuoi chay /sleep-recovery-memory-l2. null = chua tung. session-start §2.1.2 + session-end §L.b(c) doc field nay -> INFORM goi-y nen L2 neu null hoac today-last_sleep_at>=7 ngay. Lead=single-writer (chi command sleep set field nay).",
   "tiers": {
     "l1_hot": {