fix(plugins): address Codex review P2s on PR #865

1. SessionStart worker pool was sized 8 but submits up to 9 searches (3 primary
   + MAX_SHARED=6 shared) — the 9th queued behind a possibly-10s call and could
   push the hook past Claude Code's 20s SessionStart timeout before the brief
   printed. Size the pool to 3 + MAX_SHARED so nothing queues. (Introduced in
   Phase 4 when the third primary query was added.)

2. captureChattyness was dead config — written by setup but read by no hook, and
   its "heavy" level advertised "checkpoint without compaction," which no hook
   implements. The real proactivity knob is `outputStyle`. Removed
   captureChattyness entirely and folded the "how active should I be?" question
   into the output-style step (the single, actually-wired toggle); the always-on
   hooks run regardless. Updated setup skill, settings.example.json, and DESIGN.

Both flagged by Codex (review 4395911984); same class as the recallTimeframe
dead-config fix from the earlier self-review. Plugin validates; hook smoke-tested.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: phernandez <paul@basicmachines.co>

Author: phernandez

plugins/claude-code/DESIGN.md

@@ -392,7 +392,6 @@ Users get overwhelmed starting from zero. The plugin opens with a guided intervi
     "rememberFolder": "bm-remember",
     "recallTimeframe": "3d",
     "recallPrompt": "You have Basic Memory available. Search before answering recall questions. Capture material decisions inline. Cite permalinks when referencing prior work.",
-    "captureChattyness": "light",
     "preCompactCapture": "summarized",
     "placementConventions": null,
     "teamProjects": {}

plugins/claude-code/hooks/session-start.sh

@@ -141,7 +141,10 @@ RECENT_SESSIONS = ["--type", "session", "--after_date", recall_timeframe]
 # --- Run everything concurrently ---
 # Cloud reads cost a network round-trip each; parallelism keeps total wall-clock at
 # ~one query instead of the sum. Each call is independently best-effort.
-with ThreadPoolExecutor(max_workers=8) as pool:
+# Size the pool to cover every submitted search (3 primary + up to MAX_SHARED),
+# so none queues — a queued call could otherwise serialize behind a slow one and
+# push the hook past Claude Code's SessionStart timeout before the brief prints.
+with ThreadPoolExecutor(max_workers=3 + MAX_SHARED) as pool:
     fut_tasks = pool.submit(search, ACTIVE_TASKS, primary_project or None)
     fut_decisions = pool.submit(search, OPEN_DECISIONS, primary_project or None)
     fut_sessions = pool.submit(search, RECENT_SESSIONS, primary_project or None)

plugins/claude-code/settings.example.json

@@ -7,7 +7,6 @@
     "rememberFolder": "bm-remember",
     "recallTimeframe": "3d",
     "recallPrompt": "You have Basic Memory available for this project. Before answering recall questions (\"what did we decide\", \"where did we leave off\"), search the graph first — prefer structured filters (search_notes with type/status). When the user makes a material decision, capture it as a note with type: decision. Cite permalinks when referencing prior work.",
-    "captureChattyness": "light",
     "preCompactCapture": "extractive",
     "placementConventions": null,
     "teamProjects": {

plugins/claude-code/skills/setup/SKILL.md

@@ -52,14 +52,7 @@ Ask only what you can't infer. Cover:
    Keep `primaryProject` a project the user owns for their *own* capture; team
    projects are for reading and deliberate sharing only.
 
-4. **Chattiness.** "How active should I be?"
-   - **light** (default) — session brief on start, checkpoint before compaction,
-     `/basic-memory:remember` on demand.
-   - **standard** — also capture material decisions inline (enable the output style).
-   - **heavy** — also write a session checkpoint even without compaction.
-   Map the answer to `captureChattyness`.
-
-5. **Learn conventions** (optional). "Want me to look at your existing notes and
+4. **Learn conventions** (optional). "Want me to look at your existing notes and
    note your conventions so I place new notes consistently?" If yes, inspect the
    project: `list_directory` for the folder layout, sample a few notes per folder
    (and, where a folder holds recurring typed notes, you may run `schema_infer` to
@@ -68,14 +61,18 @@ Ask only what you can't infer. Cover:
    that string as `placementConventions`. Infer from their *real* notes; don't
    impose a structure.
 
-6. **Schemas.** "I'll add schemas for session checkpoints, decisions, and tasks so
+5. **Schemas.** "I'll add schemas for session checkpoints, decisions, and tasks so
    I can find them precisely later — okay?" (See "Seed the schemas" below.)
 
-7. **Output style.** "Enable the capture reflexes now? (search before recalling,
-   capture decisions as typed notes, cite permalinks)" → sets
-   `outputStyle: "basic-memory"`. Default yes for standard/heavy, ask for light.
+6. **How active should I be? (output style)** "Want me to proactively capture —
+   search the graph before recalling, write material decisions as typed notes, and
+   cite permalinks? Or keep it quiet (just the session brief, the PreCompact
+   checkpoint, and `/basic-memory:remember` on demand)?" Enabling it sets
+   `outputStyle: "basic-memory"`. Default to enabled; leave it off for a recall-only,
+   low-noise setup. (This is the single knob for how proactive the assistant is —
+   the hooks always run regardless.)
 
-8. **Shared skills** (optional, default yes). "Want the full Basic Memory toolkit —
+7. **Shared skills** (optional, default yes). "Want the full Basic Memory toolkit —
    the shared `memory-*` skills (`memory-notes`, `memory-tasks`, `memory-research`,
    `memory-schema`, `memory-defrag`, …)? I can install them alongside this plugin."
    These are the canonical, framework-agnostic skills (the same set OpenClaw bundles).
@@ -123,7 +120,6 @@ Build the `basicMemory` block from the interview:
     "captureFolder": "sessions",
     "rememberFolder": "bm-remember",
     "recallTimeframe": "3d",
-    "captureChattyness": "<light|standard|heavy>",
     "preCompactCapture": "extractive",
     "placementConventions": "<inferred summary, or null>",
     "teamProjects": {}