docs(session-memory): preserve bounded selected session scope

capabilities/families/session-memory.yaml

@@ -512,7 +512,9 @@ nodes:
     contract_level: executable
     primary_parent: session-memory.stewardship.project
     owner: {repo: aoa-session-memory, surface: skills/aoa-session-reindex/SKILL.md, authority: authored}
-    lifecycle: *deferred_lifecycle
+    lifecycle:
+      <<: *deferred_lifecycle
+      version: 1.0.1
     applicability:
       use_when: [Preserved archives require projection rebuild after source-owned classifier, schema, taxonomy, or relation changes.]
       do_not_use_when: [Raw evidence itself is missing or the request only needs a query against a current index.]
@@ -596,7 +598,9 @@ nodes:
     contract_level: executable
     primary_parent: session-memory.stewardship.curate
     owner: {repo: aoa-session-memory, surface: skills/aoa-session-batch-distill/SKILL.md, authority: authored}
-    lifecycle: *deferred_lifecycle
+    lifecycle:
+      <<: *deferred_lifecycle
+      version: 1.0.1
     applicability:
       use_when: [A reviewed set of indexed sessions needs a bounded provisional distillation wave.]
       do_not_use_when: [Exactly one indexed session is in scope; use first-pass distill., The operator expects automatic promotion.]
@@ -774,7 +778,9 @@ nodes:
     contract_level: executable
     primary_parent: session-memory.stewardship.name
     owner: {repo: aoa-session-memory, surface: skills/aoa-session-phase-discovery/SKILL.md, authority: authored}
-    lifecycle: *deferred_lifecycle
+    lifecycle:
+      <<: *deferred_lifecycle
+      version: 1.0.1
     applicability:
       use_when: [A long or multi-topic session needs phase candidates before semantic naming.]
       do_not_use_when: [A short session already has one reviewed semantic name.]

generated/capability_graph.md

@@ -2,7 +2,7 @@
 
 Derived from `capabilities/families/*.yaml`. This file is a read model, not capability authority.
 
-Source content hash: `8e0659361a4ff83f8b7bbcaea696e4dadfb41b3871bdec37b81a0825fd6a9fd7`
+Source content hash: `3aaf12c1c030d6381e8b864db4ee8397cf11648edf3e314cd8c0686ea559ef80`
 
 ## Semantic tree
 

skills/aoa-session-batch-distill/SKILL.md

@@ -23,13 +23,25 @@ truth.
 
 ## Procedure
 
+Batch distillation is set-oriented and has no positional session argument.
+Declare a bounded set with `--since`, `--since-days`, `--until`, and/or
+`--limit`, then carry the same predicate through any separately authorized
+follow-on route. The current batch and title-repair implementations use the
+same chronological registry selector, but a report is the binding set: before
+any follow-on apply, compare its selected session IDs with the prior report.
+Stop on any mismatch and re-plan that changed set under explicit scope
+authorization. An archive-wide `all` selector in a maintenance command is a
+new scope decision; it is never implied by processing one session or one
+bounded batch.
+
 Start with a planning report:
 
 ```bash
 python3 scripts/aoa_session_memory.py batch-distill \
   --workspace-root <workspace-root> \
   --aoa-root <aoa-root> \
   --since-days 21 \
+  --limit 10 \
   --write-report
 ```
 
@@ -47,13 +59,15 @@ Inspect owner quality before applying a broad pass. Each profile contains
 `project_grounding` and `owner_resolution`; fallback grounding is not the same
 as a resolved owner.
 
-Repair weak imported titles before a broad manual review wave:
+If title repair is separately authorized for this same bounded set, repair
+weak imported titles before a broad manual review wave:
 
 ```bash
 python3 scripts/aoa_session_memory.py repair-session-titles all \
   --workspace-root <workspace-root> \
   --aoa-root <aoa-root> \
   --since-days 21 \
+  --limit 10 \
   --write-report
 ```
 
@@ -66,6 +80,7 @@ python3 scripts/aoa_session_memory.py batch-distill \
   --workspace-root <workspace-root> \
   --aoa-root <aoa-root> \
   --since-days 21 \
+  --limit 10 \
   --apply \
   --write-report
 ```
@@ -93,8 +108,11 @@ promoted claims and samples, not carry the entire archive in active attention.
 
 After each batch, inspect `improvement_candidates` in the report. If the queue
 shows recurring parser misses, noisy names, missing indexes, hook gaps, or
-repeatable command patterns, improve the bundle itself with a narrow patch and
-rerun `doctor`, `audit`, and tests.
+repeatable command patterns, emit the candidate as a provisional owner-routed
+handoff. Do not patch, retrain, change hooks or CLI behavior, or rerun a repair
+inside the data-processing batch. A separately authorized owner task may
+inspect the candidate, make a narrow source change, run its owner checks, and
+start a new bounded batch after review.
 
 ## Verification
 
@@ -103,6 +121,8 @@ rerun `doctor`, `audit`, and tests.
 - All produced artifacts remain explicitly provisional.
 - Owner resolution and evidence refs are present before project-specific
   interpretation.
+- Mechanics candidates remain owner-routed handoffs; the batch does not change
+  the bundle or its processing mechanism.
 - No claim, skill change, automation, or durable lesson was promoted.
 
 ## Stop Line

skills/aoa-session-memory-global-route/references/capability-router.md

@@ -2,7 +2,7 @@
 
 Generated from owner capability contracts. This card is a retrieval read model, not procedure or proof authority.
 
-Source graph hash: `8e0659361a4ff83f8b7bbcaea696e4dadfb41b3871bdec37b81a0825fd6a9fd7`
+Source graph hash: `3aaf12c1c030d6381e8b864db4ee8397cf11648edf3e314cd8c0686ea559ef80`
 Federation: `aoa-session-memory` specializes `aoa-skills:sessions`.
 
 | skill | visibility | use when | do not use when | version | fingerprint |
@@ -13,7 +13,7 @@ Federation: `aoa-session-memory` specializes `aoa-skills:sessions`.
 | `aoa-codex-hooks-status` | deferred | Native Codex hook registration or trust must be inspected for one selected root. | The runtime is not Codex or the question concerns archive content rather than hook state. | 1.0.0 | `adc1e110880d28ee` |
 | `aoa-codex-session-segment-archive` | deferred | A concrete Codex transcript must be archived or a missed archive event must be replayed. | The source is not a Codex transcript or the raw source cannot be read and verified. | 1.0.0 | `06f718756dbb22dd` |
 | `aoa-session-archive-init` | deferred | A workspace lacks a valid session-memory root or requires an explicit install or repair. | The selected root is already healthy and no install or repair was requested. | 1.0.0 | `87fec5f9466556ab` |
-| `aoa-session-batch-distill` | deferred | A reviewed set of indexed sessions needs a bounded provisional distillation wave. | Exactly one indexed session is in scope; use first-pass distill.; The operator expects automatic promotion. | 1.0.0 | `653ff014b7b39c0f` |
+| `aoa-session-batch-distill` | deferred | A reviewed set of indexed sessions needs a bounded provisional distillation wave. | Exactly one indexed session is in scope; use first-pass distill.; The operator expects automatic promotion. | 1.0.1 | `07641eb703fa9889` |
 | `aoa-session-experience-metabolism` | deferred | A bounded set of closed; reviewed; aligned stage profiles needs privacy-safe recurring-motif candidates. | The evidence is raw; live; provisional; or asks for an eval verdict or automatic adoption. | 1.0.0 | `c7b79fab929414cd` |
 | `aoa-session-first-pass-distill` | deferred | One indexed session needs provisional candidate extraction. | The task asks for reviewed truth; final promotion; or a multi-session conveyor. | 1.0.0 | `2b577834951dd51f` |
 | `aoa-session-history-import` | deferred | Historical Codex JSONL sources must be discovered and imported into one selected root. | Only the current live transcript needs archiving or the history source is not explicitly bound. | 1.0.0 | `9aa0e6b23b0a92fa` |
@@ -23,10 +23,10 @@ Federation: `aoa-session-memory` specializes `aoa-skills:sessions`.
 | `aoa-session-memory-stress-pass` | deferred | A large archive or logical compaction boundary requires bounded stress validation. | A normal root health or release readiness check is sufficient. | 1.0.0 | `2f7b1a05b8f48775` |
 | `aoa-session-naming-readiness` | deferred | A naming wave or broad semantic renaming is being considered. | A single reviewed name is already ready for explicit application. | 1.0.0 | `fb178af39591faa8` |
 | `aoa-session-naming-wave` | deferred | A reviewed bounded set of phase or title candidates is ready for a guarded naming wave. | Readiness; phase candidates; review approval; or current index freshness is missing. | 1.0.0 | `79685864e99daf24` |
-| `aoa-session-phase-discovery` | deferred | A long or multi-topic session needs phase candidates before semantic naming. | A short session already has one reviewed semantic name. | 1.0.0 | `f42f29065ad5da68` |
+| `aoa-session-phase-discovery` | deferred | A long or multi-topic session needs phase candidates before semantic naming. | A short session already has one reviewed semantic name. | 1.0.1 | `aaba1ef4cec35f32` |
 | `aoa-session-raw-diagnostic` | deferred | A transcript read failed during hook or manual archive processing. | The raw source is readable or no concrete archive failure exists. | 1.0.0 | `d31074a22beda78e` |
 | `aoa-session-rehydrate` | deferred | A named archived session must be resumed or inspected without full raw loading. | The current active session already contains the required state or raw forensic review is explicitly required. | 1.0.0 | `dfe4d4924145df48` |
-| `aoa-session-reindex` | deferred | Preserved archives require projection rebuild after source-owned classifier; schema; taxonomy; or relation changes. | Raw evidence itself is missing or the request only needs a query against a current index. | 1.0.0 | `5749a8e863003dac` |
+| `aoa-session-reindex` | deferred | Preserved archives require projection rebuild after source-owned classifier; schema; taxonomy; or relation changes. | Raw evidence itself is missing or the request only needs a query against a current index. | 1.0.1 | `aff124cb63b3067e` |
 | `aoa-session-search` | deferred | A bounded archive query or explicit portable search-index refresh is required. | A current file; owner API; or typed entity route answers the question without archive search. | 1.0.0 | `42e8da2474fb99d3` |
 
 Load the named `SKILL.md` for procedure. Use the full generated graph for ABI, effects, failures, relations, and composition; do not infer invocation or benefit from selection alone.

skills/aoa-session-phase-discovery/SKILL.md

@@ -37,13 +37,19 @@ python3 scripts/aoa_session_memory.py phase-discovery <session-label-or-id> \
 Then refresh the indexes:
 
 ```bash
-python3 scripts/aoa_session_memory.py naming-readiness all \
+python3 scripts/aoa_session_memory.py naming-readiness <session-label-or-id> \
   --workspace-root <workspace-root> \
   --aoa-root <aoa-root> \
   --refresh-indexes \
   --write-report
 ```
 
+Keep the same session selector through this follow-on route. The refresh may
+rewrite shared naming maps as a generated dependency, but `all` is a separate
+scope choice: use it only when that scope is explicitly authorized and a
+shared index dependency makes it necessary. Discovery never widens one
+selected session into an archive-wide naming pass by itself.
+
 ## Reading The Artifact
 
 - `naming/phase-discovery.json`: machine-readable candidate list.

skills/aoa-session-reindex/SKILL.md

@@ -21,26 +21,34 @@ need to be regenerated under the current classifier.
 
 ## Procedure
 
-Start with a dry run:
+Start with a dry run for the selected session:
 
 ```bash
-python3 scripts/aoa_session_memory.py reindex-sessions all \
+python3 scripts/aoa_session_memory.py reindex-sessions <session-label-or-id> \
   --workspace-root <workspace-root> \
   --aoa-root <aoa-root> \
   --dry-run \
   --write-report
 ```
 
-For a bounded smoke pass:
+For an explicitly authorized bounded batch with a shared projection
+dependency, use `all` with the same declared window and limit in every batch
+command:
 
 ```bash
 python3 scripts/aoa_session_memory.py reindex-sessions all \
   --workspace-root <workspace-root> \
   --aoa-root <aoa-root> \
+  --since-days 21 \
   --limit 10 \
   --write-report
 ```
 
+`all` is never an automatic companion to a one-session trial. Carry the
+selected session or the exact bounded batch predicate into each follow-on
+route, and obtain separate scope plus a shared-dependency reason when moving
+from one session to `all`.
+
 For one target session:
 
 ```bash
@@ -51,12 +59,25 @@ python3 scripts/aoa_session_memory.py reindex-sessions <session-label-or-id> \
 ```
 
 After a classifier or schema reindex, refresh generated projections through the
-named catch-up route:
+named catch-up route for the same selected session:
+
+```bash
+python3 scripts/aoa_session_memory.py projection-catchup <session-label-or-id> \
+  --workspace-root <workspace-root> \
+  --aoa-root <aoa-root> \
+  --apply \
+  --write-report
+```
+
+For a separately authorized bounded batch, repeat the same `all` selector and
+window explicitly:
 
 ```bash
 python3 scripts/aoa_session_memory.py projection-catchup all \
   --workspace-root <workspace-root> \
   --aoa-root <aoa-root> \
+  --since-days 21 \
+  --limit 10 \
   --apply \
   --write-report
 ```