session 2026-07-20: /jurist-package built + PENDING-67/REVIEWED-67 + F2 ruled + v2.3.0/REVIEWED-68 operative

claude/memory/skill-harvest-register.md

@@ -8,7 +8,7 @@ metadata:
   node_type: memory
   type: reference
   originSessionId: a8eb1d46-314e-4545-a133-f747c69ad5b4
-  modified: 2026-07-19T21:06:45.514Z
+  modified: 2026-07-20T19:11:51.018Z
 permalink: claude-memory/skill-harvest-register
 ---
 
@@ -32,6 +32,18 @@ The single place proposed skills live so they don't evaporate between sessions.
 
 *Post-review additions to the Stroke-2 batch list (same authorization — "all earned ladder entries"; harvested from the pilot-ruling block later the same night):* **byte-check-hand-classifications-at-source-before-citing** (two of my classes AND the jurist's Q3 assessment overturned by bytes — the F7 lesson, jurist-directed at F3's own numbers, discharged strengthening the finding) · **coverage-metrics-are-blind-to-block-moves** (jurist-minted from their own refuted assessment: any permutation preserving local windows passes coverage; refuted-by-default unless demonstrated on a block move — the demonstrated-not-composed instrument). *No other harvest from the late-night block — the session's harvest WAS the ruled review; nothing manufactured.*
 
+## Harvest 2026-07-20 (the /jurist-package build + PENDING-67/F2/v2.3.0 arc)
+
+**BUILT this session (was authorized "build next" at the 07-19 harvest review):** **`/jurist-package`** — `~/dotfiles/claude/skills/jurist-package/SKILL.md` (symlinked into `~/.claude/skills/`; the authored-skill convention). Grounded in the 2 proven exemplars, not recall. **Proven 2× same session** (the PENDING-67 package + the F2 correction relay, both hook-passed on first write, both ruled). Absorbs `/spec-amendment` (supersession → the ladder, as ruled). `/model-handoff` remains the next authorized build.
+
+**New PATCH proposal (awaiting steward):**
+
+| Skill | Kind | One-line | Origin | Status |
+|---|---|---|---|---|
+| `/jurist-package` | patch | Add to the **after-ruling / Part-A drafting** guidance: *when transcribing a RULED finding into constitutional/amendment prose, trace the finding's own **inference direction**, not just quote its words — a faithful transcription of the WORDS can invert the INFERENCE.* Earned HARD this session, **twice**, both jurist-caught: (a) the PENDING-67 Q1 lean ("no new constitutional text") faithfully read §V's prohibitions but missed §V's *silence* on the enforcement locus — the gap itself; (b) the v2.3.0 Part-A draft wrote "re-conversion is *not* vacuous," a different sentence producing the *identical* misreading the Q2 ruling was written to prevent, inside that very subsection. The skill already names consequence-trace-one-level-deeper for the PACKAGE's own claims; this extends it to the act of **transcribing someone else's ruled finding** — a distinct, recurring failure surface. Kin to Symmetria §3 `the-say-do-seam` / `consequence-trace-one-level-shallow`, at the constitution-drafting layer. | 2026-07-20 (2× jurist-caught) | **PROPOSED** |
+
+*(One candidate, load-bearing: it recurred twice in one session and the jurist caught both. The skill carries the right discipline for its own claims but not for the transcription act; the patch closes exactly the gap I fell through — the yardstick met, the jurist had to correct the same shape twice.)*
+
 ## Open proposals
 
 *Full review with steward 2026-06-05 (housekeeping session) — every item ruled, evidence-checked against the KG drift/built record. Verdicts below.*