session 2026-08-06 evening: the parse fix landed + REVIEWED-87..94 placed + PENDING-108..111 filed

Governance: the register could not answer 'how many rulings do I owe' (23, not the
digest's 26). Seven decisions that existed only in a narrative are now placed, five
of them reconstructions carrying provenance lines. PENDING-99/-105/-106 closed (106
by split). PENDING-108/-109/-110/-111 filed.

Engine: retrieve.py accepts a sentence (27b79ca). 26 crashes -> 0, MISLOCATED 0,
FALSE-POSITIVE 0, HIT 0/22 — the engine now grounds nothing honestly, and 0/22 is
recorded as the number to beat.

PENDING-111 + jurist package: fidelity_equivalence@3 erases Alexander's invariant
rating, found by the steward reading his printed copy. Relayed for ruling.

Next session step 0, steward-directed: the MEMORY.md trim (19.9 KB vs <17.1 KB
target; relocation not deletion), then N1.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AB3Kryoy6b1pm2Nz1DYdLh

claude/memory/skill-harvest-register.md

@@ -278,3 +278,13 @@ The single place proposed skills live so they don't evaporate between sessions.
 | 184 | `/wake-up` + general | patch | **Before asking the steward to supply inputs, check whether the repo already holds them.** One level past `resurface-banked-notes-before-rederiving`: not re-deriving a banked *note* but re-sourcing banked *material*. | 2026-08-06. Was one keystroke from asking the steward to invent questions for the corpus, while `corpus/chavruta-ground-truth.yaml` held **27 real queries with audited answers** from his own hand-run chavruta. The steward caught it with four words. | PROPOSED |
 
 **Classification note:** all five `[PROPOSAL]`, none FIX-lane. 180/181/182 add ladder/flag entries but each changes what the executor must *do* before asserting (the latitude clause). 183/184 change skill procedure. Lane is provisional; when in doubt, propose.
+
+## 2026-08-06 evening — three verification-ladder entries (PROPOSED; queue with the Stroke-2 batch)
+
+*No skill created, patched or retired this session. The steward re-explained nothing a skill could have carried — what he supplied was domain knowledge from a printed book, which no skill can hold. Recording that as the honest outcome rather than manufacturing a change.*
+
+- **`a-better-than-baseline-result-earns-the-same-scrutiny-as-a-worse-one`** — the parse fix turned B11 from "returned citations" into a correct decline, which looked like the fix *removing* a false positive. Chasing why (re-running `git HEAD`'s own code) showed the improvement was not real: B11 had always declined, and yesterday's measurement doc had recorded it wrongly, along with "0 empties — the warrant machinery was never exercised." A result that flatters the change is a claim like any other. **Earned 2026-08-06; it corrected a filed measurement and regraded a pre-registered prediction from UNTESTED.**
+
+- **`read-the-consumer-before-editing-a-declarative-field`** — kin to the banked `read-the-gate's-decision-code-before-designing-its-consumer`, one level over: before changing a *declaration* (`sidecar: none-yet`), read what branches on it. Doing so converted a wrong claim ("N1 would skip two sources" — false; `chunker.load_sidecar()` reads the file and ignores the field) into the real finding: `ingest_gate.py:142` only fires "required but absent" when the declaration says `required`, so the stale value is a **disarmed tripwire** — harmless while the files exist, silent the moment one is deleted. **Census-01's decay-not-construction finding, instantiated.**
+
+- **`census-by-content-volume-not-by-marker-count`** — "240 patterns have ≥3 chunks between headings" did not establish 240 pattern *bodies*; an index or TOC produces the same signal. Re-measured by characters per span (median 4,810, zero stubs) it did. Counting markers answers a question about markup; counting content answers the question asked. **Earned twice in one exchange** — the same slip underlay reading a usage note as a bibliographic claim.