Smart-AI-Memory · silversurfer562 · May 12, 2026
diff --git a/docs/specs/archive/README.md b/docs/specs/archive/README.md
@@ -0,0 +1,36 @@
+# Archived Specs
+
+Specs that are complete, obsolete, or superseded. Moved here so
+the active `docs/specs/` directory reflects only in-flight work.
+
+The session-start hook (`plugin/hooks/spec_orient.py`) walks
+`docs/specs/` one level deep — `archive/` itself has no spec
+files at its top level, so the discovered "spec dir" returns
+`None` from `_phase_for_dir` and is silently skipped. Archived
+specs do NOT appear in session orientation.
+
+## Archived 2026-05-12
+
+| Spec | Why archived |
+|---|---|
+| `telemetry/` | Complete since v3.8.2; ongoing maintenance lives in code. |
+| `ci-debt/` | Complete 2026-05-10 across Phases A/B/C. |
+| `probe-c-memory-investigation/` | Resolved same-day with one-line `patch("threading.Thread")` fix. Lesson preserved in CLAUDE.md. |
+| `ops-specs-features/` | Shipped in three phases (PR #236 / #239 / #240). |
+| `ops-security-hardening/` | Implemented in PR #254 (`a5c50bd1`). |
+| `larger-runners/` | Closed — Probe-C resolved the OOM concern; spec itself admits the case is smaller now. |
+
+## Why archive instead of delete
+
+Git history preserves the analysis, but archive makes the
+reasoning easier to find without `git log` archaeology. Useful
+when a similar problem surfaces months later and you want to
+check whether you considered it before.
+
+## When to revisit
+
+- An archived spec is relevant again? Move it back to
+  `docs/specs/` (with a note in `decisions.md` explaining the
+  unarchive).
+- An archived spec was wrong (the resolution didn't hold)? Open
+  a new spec that supersedes it; cross-link.
diff --git a/docs/specs/ci-debt/design.md → docs/specs/archive/ci-debt/design.md b/docs/specs/ci-debt/design.md → docs/specs/archive/ci-debt/design.md
diff --git a/docs/specs/ci-debt/requirements.md → docs/specs/archive/ci-debt/requirements.md b/docs/specs/ci-debt/requirements.md → docs/specs/archive/ci-debt/requirements.md
diff --git a/docs/specs/ci-debt/tasks.md → docs/specs/archive/ci-debt/tasks.md b/docs/specs/ci-debt/tasks.md → docs/specs/archive/ci-debt/tasks.md
diff --git a/docs/specs/larger-runners/decisions.md → ...specs/archive/larger-runners/decisions.md b/docs/specs/larger-runners/decisions.md → ...specs/archive/larger-runners/decisions.md
diff --git a/docs/specs/larger-runners/tasks.md → docs/specs/archive/larger-runners/tasks.md b/docs/specs/larger-runners/tasks.md → docs/specs/archive/larger-runners/tasks.md
diff --git a/...specs/ops-security-hardening/decisions.md → ...chive/ops-security-hardening/decisions.md b/...specs/ops-security-hardening/decisions.md → ...chive/ops-security-hardening/decisions.md
@@ -1,6 +1,7 @@
 # Decisions — Ops Dashboard Security Hardening
 
-**Status:** draft
+**Status:** complete (2026-05-12) — implemented in PR #254 (squash-merge `a5c50bd1`)
+**Archived:** 2026-05-12
 **Owner:** Patrick
 **Opened:** 2026-05-11
 **Trigger:** Code-review report on PR #251 (2026-05-11) flagged the ops dashboard's command-execution endpoints as vulnerable to DNS-rebinding attacks. Verified real. Not introduced by #251 — pre-existing since the ops runner shipped.

diff --git a/docs/specs/ops-security-hardening/design.md → .../archive/ops-security-hardening/design.md b/docs/specs/ops-security-hardening/design.md → .../archive/ops-security-hardening/design.md
diff --git a/...cs/ops-security-hardening/requirements.md → ...ve/ops-security-hardening/requirements.md b/...cs/ops-security-hardening/requirements.md → ...ve/ops-security-hardening/requirements.md
diff --git a/docs/specs/ops-security-hardening/tasks.md → ...s/archive/ops-security-hardening/tasks.md b/docs/specs/ops-security-hardening/tasks.md → ...s/archive/ops-security-hardening/tasks.md
diff --git a/docs/specs/ops-specs-features/decisions.md → ...s/archive/ops-specs-features/decisions.md b/docs/specs/ops-specs-features/decisions.md → ...s/archive/ops-specs-features/decisions.md
diff --git a/docs/specs/ops-specs-features/tasks.md → ...specs/archive/ops-specs-features/tasks.md b/docs/specs/ops-specs-features/tasks.md → ...specs/archive/ops-specs-features/tasks.md
diff --git a/...probe-c-memory-investigation/decisions.md → ...probe-c-memory-investigation/decisions.md b/...probe-c-memory-investigation/decisions.md → ...probe-c-memory-investigation/decisions.md
diff --git a/...ecs/probe-c-memory-investigation/tasks.md → ...ive/probe-c-memory-investigation/tasks.md b/...ecs/probe-c-memory-investigation/tasks.md → ...ive/probe-c-memory-investigation/tasks.md
diff --git a/docs/specs/telemetry/design.md → docs/specs/archive/telemetry/design.md b/docs/specs/telemetry/design.md → docs/specs/archive/telemetry/design.md
diff --git a/docs/specs/telemetry/requirements.md → docs/specs/archive/telemetry/requirements.md b/docs/specs/telemetry/requirements.md → docs/specs/archive/telemetry/requirements.md
diff --git a/docs/specs/telemetry/tasks.md → docs/specs/archive/telemetry/tasks.md b/docs/specs/telemetry/tasks.md → docs/specs/archive/telemetry/tasks.md
diff --git a/docs/specs/redis-decoupling/decisions.md b/docs/specs/redis-decoupling/decisions.md
@@ -1,10 +1,30 @@
 # Per-decision log — Redis decoupling
 
+**Status:** GATED (2026-05-12 grooming pass)
+
 Append-only log. Pre-flight findings (Phase 3A tasks #1–#2) below.
 Per-phase execution decisions appended as commits land.
 
 ---
 
+## Gate decision (2026-05-12)
+
+This spec is approved (Phase 3A done) but deliberately gated.
+**Don't implement until:** the memory backend abstraction shows
+real friction in production use — e.g., a user reports they want
+to swap Redis for another KV store, or a feature gets blocked by
+the current Redis coupling. Today, the coupling isn't blocking
+anything that's been asked for.
+
+If the trigger fires, pick this spec up from where Phase 3A left
+off; the analysis is still load-bearing.
+
+If the trigger never fires within ~6 months (by 2026-11-12),
+revisit whether the spec itself is still relevant or should move
+to archive.
+
+---
+
 ## Phase 3A pre-flight (2026-05-10)
 
 ### Task #1 — `attune_redis` PyPI state → forces **C1** (deprecate + delete)