1.5.8

docs/GO-LIVE-v1.5.0.md

@@ -51,3 +51,8 @@ Für einen realen Research→Staging-Smoke-Test zusätzlich verifizieren:
 ## v1.5.7 Identifier-Grounding Zusatzgate
 
 Bei Windows-/Vendor-Artikeln dürfen normale Slash-Komposita oder URL-Pfade kein `source-unverified identifiers` auslösen. Echte CLI-Switches in Code-Spans/Fences bleiben source-verifiziert. Vor Go-Live mindestens einen Goal-Lauf mit `BIOS-/UEFI`-ähnlicher Prosa und einen Lauf mit einem belegten Slash-Command prüfen.
+
+
+## v1.5.8 Staging-Revalidation Zusatzgate
+
+Nach einem Upgrade auf v1.5.8 aktive Goals mindestens einen Scheduler-Cycle durchlaufen lassen. Ein Artikelziel darf erst dann als erfüllt gelten, wenn der Goal-State `staging_draft_validated=true` und `staging_quality_gate_version=staging-v3` meldet. Persistierte Fehler aus älteren Gates dürfen nicht als aktueller Fehlerstatus interpretiert werden.

docs/MIGRATION-v1.5.7-to-v1.5.8.md

@@ -0,0 +1,16 @@
+# Migration v1.5.7 → v1.5.8
+
+v1.5.8 is a non-destructive goal/staging state migration. No volumes must be deleted.
+
+1. Build/publish the v1.5.8 images.
+2. Set `IMAGE_TAG=1.5.8`.
+3. Pull and recreate NeuroForge and its worker:
+
+```bash
+docker compose --profile research pull neuroforge neuroforge-worker
+docker compose --profile research up -d --force-recreate neuroforge neuroforge-worker
+```
+
+4. Keep all existing goals and volumes.
+5. Let each active goal run once. Existing legacy drafts will either be revalidated under the current gate, remain below the current evidence threshold, or receive a current quality error.
+6. Article goals should only return to 100% after `staging_draft_validated=true`.

docs/RELEASE-v1.5.8.md

@@ -0,0 +1,20 @@
+# GLPI NeuroForge Mega v1.5.8
+
+v1.5.8 closes a production-state consistency bug found while revalidating existing FortiClient, DISM and BitLocker goals after the v1.5.7 identifier-grounding fix. Older `last_staging_error` values could remain visible forever because a goal with an existing draft returned early whenever the newest research run contributed no new evidence. The old draft also continued to satisfy an article target even if it had never passed the current quality gate.
+
+## Staging revalidation
+
+- Goals persist `staging_draft_validated`, `staging_quality_gate_version` and a hashed `last_staging_attempt_signature`.
+- Existing drafts from an older gate are revalidated once against the current gate using already persisted relevant evidence.
+- Unchanged, already validated drafts are not regenerated on empty scheduler cycles.
+- Unchanged deterministic failures are not retried every cycle; the evidence signature prevents expensive LLM retry loops.
+- Transient transport/publisher failures remain retryable.
+- If current evidence no longer satisfies the quality threshold, the old draft is marked not validated and the stale error is replaced by the current quality state.
+
+## Goal progress semantics
+
+For article/draft targets, historical `staging_drafts_created` is audit history only. Progress is now based on whether the current stable staging draft exists and is validated under the current quality-gate version. This prevents `1/1 = 100%` for legacy or invalidated drafts.
+
+## Compatibility
+
+No destructive data migration is required. New fields are optional JSON fields and are populated by the next goal cycle. Existing goals, memories, synapses, research runs and staging drafts are retained.