session 2026-06-05 evening: housekeeping wrap — AI-posture ruling memory, MemPalace forensics + workaround, harvest register ruled, verification ladder, Symmetria flags, register wiring applied

claude/memory/session-ledger-2026-06-05.md

@@ -36,3 +36,10 @@ metadata:
 
 ## Bypasses
 - 2026-06-05T17:25 — CHAMBER-LIBRARY LFS RETIRED + PUSHED (housekeeping item 1). Root cause of the 7-day block: Gitea has no LFS endpoint (batch 404) AND origin/main held pointer stubs since the initial import — the remote was never a backup of the corpus. Steward decided: retire LFS (markdown is git-native; LFS stores full blobs per version — wrong shape for cruft-removal + studium-markup futures). Executed: APFS-clone backup → commit 8 waiting texts + 2 scripts (f851336) → git lfs migrate export across all 17 commits → .gitattributes removed (0677e8a) → gc 1.4G→353M → force-push (safe: remote had stubs only, no other clones) → FRESH-CLONE GATE: 407 files / 668M / zero pointer stubs. Residue surfaced: pre-rewrite backup dir (steward to delete) + CHAMBER_CATALOGUE.md stale since March.
+- 2026-06-05T18:40 — WRAP/WAKE DOTFILES-PUSH PATCHES BUILT + FIRST LIVE RUN (housekeeping item 2). Steward proposed rolling the dotfiles push out of sysupdate into the session skills; authorized conversationally. wrap-up §6.5 (scoped add — never add-., session-stamped message, preservation-not-modification boundary, fail-loudly) + wake-up §2.c read-only push-check (wake reads, never mutates). Register records lineage. First run: 8cca9de — REVIEWED-28/29 (steward-filed from executor drafts) + both skill patches + 6 memory files pushed; dotfiles level with remote. Governance-state single-disk latency closed at the session boundary.
+- 2026-06-05T19:05 — RETURN: censused-routes drift, 3rd instance TODAY, same shape each time — `git status | head -N` truncation read as the full census (CM-AI '2 untracked' was 8; studium '2' was 7; the studium git-add swept 5 unnamed-but-belonging files into 8081b12 — substance right, message under-describes). The standing recalibration said exactly this; caught at commit-output not before. Rule going forward: NEVER pipe a census through head — count first (wc -l), then look.
+- 2026-06-05T19:00 — CM-AI + STUDIUM TRACKED & PUSHED (housekeeping item 3, steward 'yes to all'). CM-AI: 8 artifacts (REVIEWED-S0-cited methodology trio = constitutional lineage off-disk at last; lens cards path-pinned by scan skills; studium scan-register relocation = NOTED DEFERRAL for a studium session). Push rejected → SEB MOVED: 12 commits 05-25→06-03 (A2 benchmark-governance v1.0, epistemic-gates L1 spec v1.0, Amendment 61 parent, B1.3 instrumentation, COVER NOTE FOR DAVID b4daa71, plan-010 Kronos×Plane, memory-os + PMB deep-reads); zero overlap, rebased, pushed f3beead. Studium: 7 docs incl. jurist settlements, pushed 8081b12.
+- 2026-06-05T19:20 — STEWARD DECISION: Seb cover-note reply (b4daa71, 2026-06-02) DEFERRED TO TOMORROW AS HARD COMMITMENT — 'we must do this tomorrow.' Scope agreed: Tier A + probe (~2–2.5h) — #4 emphatic (A2.4 degraded-env refusal = honest-degradation doctrine), #1 doctrinal half (the hang IS the bug; BOOTSTRAPPING must refuse loudly, not wedge cursors), #3 A2-invariants read, Pebbles convergence line, + mindfabric-00 graduation probe (#6, answers #5's premise). Drafted in Seb's format (David + Claude), steward reviews before send. Tier B code-grounding rides a second note. THIS SUPERSEDES bio/CV as tomorrow's opening priority — 'parked till mid-July' framing is STALE (Seb is moving: 12 commits, six prioritized asks).
+- 2026-06-05T19:45 — SYMMETRIA CHECK: CloudFlare AI-posture ruling (slate item 4). NEW FINDING during verification: beyond the robots.txt signal, CloudFlare WAF is ENFORCING — Claude-User/Anthropic-infra fetches of ARC pages → 403 (plain curl + Googlebot-UA → 200, tested twice). ARC is currently unreadable by answer-time AI retrieval — readers' AI assistants cannot quote/cite the live text; representation degrades to training echoes (anti-provenance). Signal layer verified more surgical than recorded: search=yes, ai-train=no (EU Art.4 reservation language), ai-input UNSET, disallows = training crawlers specifically. Executor opinion given with conflict-of-interest named (AI advising on AI-training opt-out; Anthropic bot on the list): ratify ai-train=no, LIFT the WAF reader-path block, set ai-input explicitly, move authorship dashboard→repo+spec. Steward ruling awaited.
+- 2026-06-05T20:30 — AI POSTURE RULING EXECUTED + LIVE (slate item 4 CLOSED). Steward ruled: ai-train=no (Art.4 reservation) + ai-input=yes + search=yes; influence-welcome-extraction-not. Four artifacts (60011fb, deployed both remotes): repo-authored static/robots.txt w/ Content-Signal line + 9 training disallows; apparatus §SEO 'AI posture' subsection (authorship rule load-bearing: dashboard never again speaks for the house); drafts/ai-posture-ruling-2026-06-05.md (full reasoning incl. conflict-of-interest note); Jekyll-era root robots.txt deleted. Dashboard: steward turned managed prepend OFF + lifted WAF AI-block (guided live; iterate-and-probe loop). Verified: signals line live, prepend 0, Claude-User 200, genuine AI-infra fetch reads /dwellings. WAF custom-rule option explained, not urged; if adopted → spec record.
+- 2026-06-05T20:55 — WAF BELT-AND-BRACES ADOPTED (steward chose after honest-marginal-value framing; Bytespider robots-ignoring = the real target). Toggle-built rule; probe matrix: 6 fetching training UAs 403, reader/search/plain 200; Google-Extended+Applebot-Extended correctly ABSENT (robots tokens, not fetching UAs — my pasted expression had over-specified, toggle list surfaced it); CCBot blocked under another label. Spec record appended to apparatus §SEO (5f-commit), pushed both remotes. Docs-only — no rebuild needed.