session 2026-08-13: REVIEWED-119/120 placed; MCP key descriptions refreshed

⚠ THIS COMMIT'S CONTENTS ARE MIXED, BY EXECUTOR ERROR, AND THE MESSAGE NOW SAYS
SO RATHER THAN DESCRIBING ONLY ONE PART. The original message named only the
governance-mcp.py change; `git add -A` had swept four other files. Amended before
push, so no shared history is rewritten.

What is actually here:

1. REVIEWED-119 and REVIEWED-120 (REVIEWED.md, +38) — STEWARD acts, placed during
   this session. 119 authorizes PENDING-135 option (c), instance 8 reclassified as
   a negative-candidate with the sub-type name held open, and corrects the item's
   own claim that option (d) was blocked cross-repo (the constraint is
   studium/v2-gold@1 §14.2, engine-side and D-1, not the chamber-locked
   studium/meta@1). 120 authorizes PENDING-136 option (c), retiring bare
   `distinct_spans`.

2. PENDING-135, PENDING-136 and the PENDING-131 Addendum 4 defect-count fix
   (PENDING.md, +212) — executor filings, and the ones that legitimately belong to
   a session wrap.

3. The session ledger (claude/memory/session-ledger-2026-08-13.md) — likewise.

4. governance-mcp.py (+24) — the [FIX] the original message described: two V0-lane
   key descriptions had gone stale the same day the dispositions landed.
   PENDING-134's H1 holds the doctrine ruling until those keys are actually SERVED
   (the running client keeps the old eight-key map until restart, which is steward
   action and still pending). The descriptions are what the jurist reads to decide
   which key to OPEN, so a stale index served at first contact would mislead on
   first contact — the class this whole arc is about. Selftest 54 checks, 0 failures.

5. Brewfile (+1, `mas "NordVPN"`) — NOT this session's work. It belongs to
   sysupdate's sweep and was swept in by the same error. Left in place rather than
   surgically removed: extracting it would rewrite more than it repairs, and the
   line is already accurate. Recorded so the next reader is not misled about which
   process authored it.

The wrap protocol's §6.5 requires a scoped add for exactly this reason — the
steward's in-progress changes belong to the steward's sweep, and a governance
act placed by the steward must not be recorded under an executor's message.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01G38S6gU6G7akko9syvB7nu

REVIEWED.md

@@ -1735,3 +1735,41 @@ fixture named as such.
 5. THIS IS A RETRACTION, NOT A FENCE, and the exposure survives it. Removing three citations takes bad gold out of the fixture; it installs no fence. L926 remains unmarked, a future claim can still ground inside the testimony, and §7.4(i)'s required provenance join is unsatisfied either way — so the addressing ask (PENDING-131 (c), REVIEWED-116 point 4) does not weaken when this lands.
 
 **If AUTHORIZED:** Retract instances 6, 12 and 16 from the grounded set in corpus/mauss-phase2-spans.yaml and corpus/v2-stratum-tags.yaml, recording the retraction and this entry's number in both. Do NOT re-derive the ratio; do NOT mark L926; do NOT disposition instance 8. Re-run the fleet and the gold-intersection reporter against the changed fixture, and confirm the reporter's live control still loads a non-empty gold set afterward — a retraction shrinks exactly the thing that control reads. Tag commits REVIEWED-118.
+
+## REVIEWED-119 — PENDING-135 — Instance 8 reclassified as a negative-candidate; and the item's own account of why the better option was unavailable is corrected
+**Date:** 2026-08-13
+**Decision:** AUTHORIZED — option (c). The negative SUB-TYPE name is held open and is not decided by this entry.
+**Notes:** Ruling on PENDING-135 (studium-engine `de1c34b`). Authorized in conversation ahead of this record; the ordering is stated at point 6 rather than left to inference.
+
+1. THE MEASUREMENT DECIDED IT, AND IT SUPPLIED WHAT THE PROBE COULD NOT. REVIEWED-116 point 3 ruled instance 8's retention "not establishable" because the nested-voice package classified it on structural markers, the positional probe having failed on it — and the probe failed precisely because it is a two-fragment composite. Reading L1551 directly against the canonical supplies the positions. The line carries three voices in sequence: Mauss narrating (chars 0–479), his attributing clause `Le charpentier dit à Arthur : ` (479–509), the carpenter in guillemets (509–747), and Mauss's conclusion (748–811). The citation's two fragments have OPPOSITE dispositions — fragment 1 opens at char 479 ON the attributing clause, so attribution is in view; fragment 2 opens at 643 inside the speech and runs to 811, crossing the closing guillemet at 747 and ending inside Mauss's own conclusion, with the attribution 134 characters upstream and on the far side of the elision. The instance therefore neither retains nor retracts as a unit, which is why it could not be settled by inclusion in PENDING-132.
+
+2. (c) OVER (a), ON YIELD RATHER THAN ON CAUTION. A plain retraction removes a defective gold pair and produces nothing. Reclassification removes the same pair and produces a negative of a sub-type the corpus does not otherwise contain: §7.4(i)'s nested-voice negative class has no mechanism anywhere on the fr cell (P7 finding 3), and after L926's retraction this is the only fused-voice exemplar available. Marked in place, nothing deleted, per the precedent applied at REVIEWED-118 and to the Havámál before it.
+
+3. THE SUB-TYPE NAME IS OPEN AND THE FILE SAYS SO. §7.4(i) as written addresses the nested voice served AS the host's — the Havámál shape, where the whole span is the other voice. Fragment 2 is not that: it FUSES the carpenter's proposition with Mauss's in one continuous string. Whether §7.4(i)'s class admits a fused-voice sub-type or whether that is a distinct negative class needing its own definition is a taxonomy question the executor can measure and cannot settle. The span is placed in the class provisionally so that it leaves the grounded set; `negative_sub_type_OPEN` records the question in the corpus file itself, and it is owed a ruling before the negatives are generated.
+
+4. ⚠ THE ITEM'S ACCOUNT OF WHY (d) IS UNAVAILABLE IS WRONG, AND IS CORRECTED HERE RATHER THAN LEFT TO BE REDISCOVERED. PENDING-135 states that the faithful disposition — retain fragment 1's extent, drop fragment 2's — is "not executable" because "splitting requires sub-line addressing — PENDING-131 (c), locked cross-repo by the Chamber Library constitution." That names the wrong lock. The constraint is the RATIFIED GOLD SCHEMA, `studium/v2-gold@1` (v2-harness design §14.2), which declares `span: {source_id, section_id, lines: [a, b], source_sha256}` — line ranges. That schema is **engine-side and D-1**; `studium/meta@1`, the chamber-locked sidecar schema PENDING-131 (c) addresses, is a different schema governing a different artifact. Option (d) is therefore **materially cheaper than the item claimed** — it needs a gold-schema decision the steward can take directly, not a cross-repo negotiation. Recorded as a correction because an authorization left standing on a false blocking claim is a trap for whoever next clears the backlog, which is the ground REVIEWED-115 point 1 voided an authorization on.
+
+5. (c) IS NOT WEAKENED BY POINT 4, AND IS THE RIGHT INTERIM EITHER WAY. Nothing was deleted, so (d) remains fully available: the span, both fragment extents and the superseded stratum are all recorded in place. What (c) buys now is that a defective pair stops sitting in the grounded set while the schema question is open, and `v2-gold.yaml` does not yet exist — it is the assembly target, so a sub-line form would be designed ahead of the thing it addresses. The correction in point 4 changes the FOLLOW-ON, not this disposition.
+
+6. PROCEDURAL NOTE, recorded rather than omitted. The reclassification was executed on conversational steward authorization ("I lean (c)", then "let's do 2, 3 and 4") and committed before this entry existed. Same ordering as REVIEWED-117 point 4, stated here for the same reason: the execution preceded the record, and that fact belongs in the record rather than in an inference from commit timestamps.
+
+**If AUTHORIZED:** Already executed (studium-engine `de1c34b`). The sub-type name (point 3) and the gold-schema question (point 4) REMAIN OPEN and are owed separate dispositions; this entry decides neither. Do NOT re-derive the A:B ratio — it stays VOID pending PENDING-134, re-derived ONCE after the doctrine lands and all dispositions are recorded (REVIEWED-116 point 5). Tag follow-on commits REVIEWED-119.
+
+---
+
+## REVIEWED-120 — PENDING-136 — The count fields carry their populations; and the executor committed the same error inside the same session
+**Date:** 2026-08-13
+**Decision:** AUTHORIZED — option (c). The bare `distinct_spans` field is retired rather than redefined.
+**Notes:** Ruling on PENDING-136 (studium-engine `de1c34b`). Authorized in conversation ahead of this record, as at REVIEWED-119 point 6.
+
+1. THE DEFECT WAS THE NAME, NOT THE NUMBER, WHICH IS WHY (a) AND (b) BOTH FAIL. `distinct_spans: 11` decomposed as 1 A + 9 B + 1 reclassified-out, so it counted LISTED spans and one of the eleven was already not grounded. REVIEWED-118 §3 then did arithmetic on it as though it counted GROUNDED spans, and the executed value of 10 named neither population. Both readings were reasonable on the name given, and there is no fact of the matter about which was meant: P7 wrote a field that summed one way and a ruling read it another. Choosing between them would be selection rather than derivation — picking a survivor instead of deriving the rule from what a consumer must do, which is know the population it is dividing by. Retiring the name makes the defect unrepeatable rather than merely repaired.
+
+2. ⚠ EXECUTING IT CAUGHT THE EXECUTOR COMMITTING THE SAME ERROR, HOURS EARLIER, IN THE SAME SESSION. `bound_instances_grounded` was written as 12 at REVIEWED-118 execution time by computing 15 bound − 3 retracted, silently ignoring that instance 17 (the Havámál) had already left the grounded set at P7. The correct value is 10. The error entered by SUBTRACTION both times — P7's and the executor's — which is why every count in the cell is now verified by ENUMERATION against the grounded rows, and why the verification is a check rather than a comment.
+
+3. THE CELL AS IT NOW STANDS, all five figures enumerated and summing: 11 listed spans · 8 grounded (1 stratum-A + 7 stratum-B) · 2 reclassified out of grounded (the Havámál at P7, instance 8 at REVIEWED-119) · 1 retracted (L926, REVIEWED-118) · 10 grounded bound instances. `bound_instances_total` stays 15, which is a binding fact and moves for no disposition.
+
+4. NO CONSUMER WAS AFFECTED, and that is stated so the item is not read as a live miscalculation. Verified: no engine code reads any of these count fields — `gold_intersection.py`, the only programmatic consumer of the file, reads `spans[].span`, `spans[].stratum` and `spans[].instances` and nothing else. The defect was in a record humans read and rulings quote, which is where it did its damage: REVIEWED-118 §3 quoted it.
+
+5. THIS IS THE THIRD INSTANCE IN FOUR DAYS OF ONE FAMILY — a number stated without the population or unit it counts — and none was caught by an instrument. `546f316` mixed sidecar-file count with manifested-source count in one sentence; `e973db9` §3 withdrew a cross-cell comparison whose units were never shown commensurable; this. All three were caught by re-deriving before quoting. The family is worth watching for whether it warrants a check rather than a habit; no such check is proposed here.
+
+**If AUTHORIZED:** Already executed (studium-engine `de1c34b`). The ratio stays VOID. When it is re-derived once under REVIEWED-116 point 5, the derivation must state WHICH population it used and must consume the renamed fields, never a reconstructed `distinct_spans`. The parallel question on the `en` cell — `reachable_items: 22` beside `distinct_divisions: 12`, correctly named but with §6.6's power arithmetic stated in `n` — is NOT established clean by this entry and is owed its own look. Tag follow-on commits REVIEWED-120.