fix(harness): remediation difetti review — gate↔CLI, D15, D7/D6, D14, robustezza

Implementazione del piano di remediation progressiva sui difetti emersi
dall'analisi dell'harness. Tutto verificato: 214 test Python (incl. L0 su
Postgres reale), 14 test JS del gate, ruff pulito.

Blocco 1 (CRITICA, integrazione gate↔CLI):
- phase advance: gate usa --auto + exit 6; reviewer_confirm kind:phase fa
  advance esplicito che applica i prerequisiti (prima non avanzava per le
  fasi a conferma umana).
- cte plan riceve i --name dal gate (param names); set-question con id
  posizionale; skill `tht search find`; nuovo comando `tht memory save-one`
  con dedup hash client-side in save_one_memory.

Blocco 2 (D15, stato post-rollback):
- campo `phase` su DecisionRecord + effective_decisions phase-aware per i
  subject "a nome" (cte_approved ecc.); _compute_promotions e finalize sulla
  vista effective; finalize confronta col piano CTE effettivo, non glob;
  `decision add --retracts` + comando `decision retract`.

Blocco 3 (D7 read-only + D6 manifest):
- assert_read_only su tutti e quattro i codepath (direct + REST);
- manifest author/summary/updated_at/updated_by/schema_version popolati +
  helper touch_manifest sulle mutazioni.

Blocco 4-5 (D14a/D14b):
- decision_min_phase data-driven via `emits:` in workflow.yaml;
- formula evidence: status auto, search_formulas, gruppo CLI `tht formula`,
  `search find --kind formula`, load_evidence_dir salta i .sql.md.

Blocco 6 (robustezza):
- taskdoc slice promoted_tables + bound enforced; report escaping/bound +
  rsplit note; filtro kind reader REST/direct; conteggio upserted robusto;
  guard REST run_query non-list; LSH disallineato -> LshIndexError.

Blocco 7 (pulizia):
- dead code gate e KIND_TO_TABLE morto rimossi; doc Postgres-only
  (README + connection.py).

Blocco 0 (parziale): test di compatibilità firma gate↔CLI
(tests/integration). Rinviati: fake-Pi runtime completo, artifact-gate da
disco (#23), parità eligibility REST/direct (#28), unificazione
reserved-labels (#30), memory_rejected da deselezione (#33).

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

harness/tht/memory.py

@@ -6,7 +6,7 @@ from pathlib import Path
 
 from pydantic import BaseModel
 
-from tht.decisions import DecisionRecord, DecisionType, list_decisions
+from tht.decisions import DecisionRecord, DecisionType
 from tht.session.models import SessionManifest
 from tht.vectorstore.records import VectorRecord
 
@@ -132,7 +132,10 @@ def _compute_promotions(
     session_dir: Path, manifest: SessionManifest, *,
     seqs: list[int] | None, existing: list[MemoryRecord],
 ) -> list[MemoryRecord]:
-    decisions = list_decisions(session_dir)
+    # Vista effective (D15, §4.8): non promuovere decisioni stale dopo un rollback.
+    from tht.phase import effective_decisions
+
+    decisions = effective_decisions(session_dir)
     selected = decisions if seqs is None else [d for d in decisions if d.seq in seqs]
     already = {(r.session_id, r.decision_seq) for r in existing}
     n = _next_id_num(existing)
@@ -234,7 +237,12 @@ def save_one_memory(
     """Targeted one-row upsert of a promoted decision to pgvector via the writer key
     (spec D11). This is NOT a full vectorstore resync: it embeds and pushes a single
     record, so a workstation with a writer key can publish one memory without
-    rebuilding the index. Returns the upsert count (0 if no record matched).
+    rebuilding the index. Returns the upsert count (0 if no record matched or the
+    record is already up to date).
+
+    Hash dedup client-side (spec §5.4): the SHA-256 of the content is compared with
+    the writer's existing_vector_hashes; the embedding (Ollama round-trip) and the
+    upsert are skipped when the content is unchanged. Idempotent by construction.
 
     `writer` is a VectorRestClient (writer key); `embedder` an embeddings client.
     The destructive cleanup (sync's delete-stale step) is intentionally absent: it
@@ -246,11 +254,15 @@ def save_one_memory(
     record = memory_vector_record_for_decision(records, decision_seq)
     if record is None:
         return 0
+    new_hash = content_hash(record.content)
+    existing = writer.existing_hashes("memory", ["memory"])
+    if existing.get(record.id) == new_hash:
+        return 0  # unchanged: skip embedding + upsert
     embedding = embedder.embed_documents([record.content])[0]
     row = {
         "record_key": record.id,
         "kind": record.kind,
-        "content_hash": content_hash(record.content),
+        "content_hash": new_hash,
         "metadata": pack_metadata(record),
         "embedding": embedding,
     }