diff --git a/README.md b/README.md
index 9d4de259..25e40510 100644
--- a/README.md
+++ b/README.md
@@ -23,26 +23,26 @@ An operating system for AI agents. Not a chatbot wrapper. Not a prompt chain. A
 
 ## Current State: Beta
 
-**It works.** 14 of 15 branches are operational, tested, and communicating. 53 PRs merged. 30+ orchestration sessions. We're past prototyping and into hardening — building the infrastructure that makes the system reliable at scale.
+**It works.** All 15 branches operational. 95+ PRs merged. 38 orchestration sessions. 733 tests across the system. 173 drone commands discovered. The system is past prototyping — we're in the hardening phase, building diagnostic tooling and iterating branch by branch.
 
 **Recently completed:**
 
-- **CLI front door + init anywhere** — full discovery chain rebuilt so users can find and use `aipass init` through natural exploration. CLI rewritten to seedgo's auto-discovery pattern. Registered as a drone internal module — `drone @cli aipass init` works from any directory on any system, before any project exists. Creates a registry (UUID), passport, `.trinity/` memories, prompt files, and hooks. Every project is fully self-contained — no shared state, no cross-contamination between workspaces.
-- **Memory introspection overhaul** — memory branch rewired to seedgo-compliant CLI discovery. Rollover and search introspection rebuilt. 18+ files ported back from archives (symbolic reasoning, templates, pool processor). Templates modernization in progress (FPLAN-0052).
-- **Seedgo standards enforcement** — checklist command fixed, checker false positives addressed, handler guards added. PR #52 open.
-- **Credential model** — UUID-based registry matching is live. Every project gets a unique registry ID, every citizen's passport carries it. `aipass init` creates a new project in any directory with its own credentials.
-- **Stderr routing** — system-wide migration across 10 branches (48 files). CLI owns the display layer (`error()`, `warning()`, `fatal()` all route to stderr). Seedgo enforces it with an automated checker.
-- **Dashboard pipeline** — Prax owns the dashboard end-to-end. Per-branch `STATUS.local.md` files sync to a central `STATUS.md` via `drone @prax status sync`.
-- **Drone test suite** — 188 tests across 5 files. Interactive mode for human-facing commands. Credential verification with dedicated error hierarchy.
-- **System governance** — git workflow, commit signing (`Co-Authored-By: @branch`), DPLAN/FPLAN documentation, and "How to Work" guidelines all codified in the global prompt.
+- **Dispatch UX redesign** — `drone @ai_mail dispatch @target "Subject" "Body"` sends + wakes in one command. `--fresh` flag for clean sessions. `email` command for mail-only (no wake). Fully tested.
+- **PR v2 workflow** — commit-on-main architecture. Changes never leave your working tree. Feature branches are just pointers for GitHub's PR system. No more disappearing files.
+- **Handler guard fix** — cross-branch handler imports blocked by `.py` files from other branches. Command-line `python3 -c` allowed through. 13 branches updated.
+- **Memory vectorization fix** — batch processing (2 subprocess calls instead of 228), decoupled from startup trigger, explicit `drone @memory process-plans` command. 113 plan files vectorized in ~1 minute.
+- **Diagnostic tooling** — 4 scanners built: dead code (14 unused files found), command inventory (173 commands), prompt quality (6 rich/4 basic/5 stub), test coverage (733 tests, 26% module coverage).
+- **Prax monitor** — fully operational with inotify file watching, branch detection, polling fallback with actionable error messages.
+- **Plan cleanup** — 60+ FPLANs/DPLANs closed. Templates updated with "close immediately when done" rule.
+- **System governance** — git workflow, commit signing, DPLAN/FPLAN documentation, logging/debugging guidelines, `.archive/` pattern all codified in the global prompt.
 
 **What we're solving now:**
 
-- **Memory subsystem wiring** — symbolic reasoning, templates, and pool processor modules ported back but need full integration and testing. Templates modernization (spawn handler, v2 schema) in progress.
-- **Lint cleanup** — 474 ruff violations remain (mostly unused imports from unwired modules). Blocked until seedgo audit coverage is higher — we can't confidently remove imports until we know what's actually used.
-- **Dashboard CLI routing** — the Python API works but `drone @prax dashboard refresh --all` fails because argparse eats the flags before the module sees them. Known bug, fix pending.
-- **Cross-platform reliability** — Linux and Windows tested. macOS is structurally supported but needs a dedicated testing pass. All paths use `pathlib`, no hardcoded paths, secrets stored at `~/.secrets/aipass/`.
-- **Agent agnosticism** — currently focused on [Claude Code](https://docs.anthropic.com/en/docs/claude-code) (hooks for auto-diagnostics, prompt injection, session recovery). But AIPass is designed to not depend on any single provider. `agents.md` and `gemini.md` can bootstrap the system for Codex and Gemini — you lose hooks but keep the core. The plan is truly agnostic: any agent, anywhere, persistent memory, no vendor lock-in.
+- **Branch-by-branch audit** — walking through every branch from devpulse, testing commands, noting issues, dispatching fixes. API branch audit in progress (DPLAN-0029).
+- **Local prompt enrichment** — 5 branches still on 14-line stubs (ai_mail, backup, cli, drone, prax). Rich prompts = less babysitting.
+- **Test coverage expansion** — 9 branches have zero tests. Building toward comprehensive coverage using the test scanner for visibility.
+- **Cross-platform reliability** — Linux and Windows tested. macOS structurally supported. All paths use `pathlib`, secrets at `~/.secrets/aipass/`.
+- **Agent agnosticism** — currently focused on [Claude Code](https://docs.anthropic.com/en/docs/claude-code) (hooks for auto-diagnostics, prompt injection, session recovery). But AIPass is designed to not depend on any single provider. `agents.md` and `gemini.md` can bootstrap the system for Codex and Gemini — you lose hooks but keep the core.
 
 ## Getting Started
 
@@ -92,7 +92,7 @@ Every branch is a citizen — an expert in its domain with its own memories and
 |--------|------|
 | `devpulse` | **Start here.** Orchestration hub — coordinates everything |
 | `drone` | AI-friendly CLI — every command is a single-line, non-interactive call |
-| `seedgo` | Standards enforcement — 24-standard audit pack, system compliance |
+| `seedgo` | Standards enforcement — 21-standard audit pack, system compliance |
 | `prax` | Logging and monitoring (the only logger in the system) |
 | `cli` | Terminal display, stderr routing, project commands |
 | `flow` | Workflow management — FPLANs (execution) and DPLANs (design) |
diff --git a/STATUS.md b/STATUS.md
index 55a53419..75c85915 100644
--- a/STATUS.md
+++ b/STATUS.md
@@ -2,19 +2,19 @@
 
 > Auto-generated by `drone @prax status sync`. Do not edit manually.
 
-**Last sync:** 2026-03-17 19:57  
+**Last sync:** 2026-03-20 00:00  
 **Summary:** 9 operational | 0 in-progress | 0 not started
 
 ---
 
-<details><summary><strong>@ai_mail</strong> — Operational (2026-03-17)</summary>
+<details><summary><strong>@ai_mail</strong> — Operational (2026-03-17 (session 14))</summary>
 
 # @ai_mail
 
 > Inter-branch email, dispatch, wake, bounce
 
 **State:** Operational
-**Last update:** 2026-03-17
+**Last update:** 2026-03-17 (session 14)
 
 ## Milestones
 - Send/receive/reply/close lifecycle
@@ -24,19 +24,22 @@
 - notify-send to dbus notifications
 - mailbox_path bug fixed (relative→absolute in get_user_by_email, get_all_users)
 - Health audit: 7 dead handlers archived, branch cleaned
+- Seedgo compliance: 93% → 99% (json_structure 100%, introspection 100%)
 
 ## Current Work
-- 1 inbox email from @api (unit test setup advice) — pending reply
+- No active tasks. Inbox clear.
 
 ## Known Issues
 - _find_repo_root() duplicated 9 times across handlers (consolidation candidate)
 - get_all_branches() has 2 implementations (delivery.py vs registry/read.py)
 - Registry shows 6 pytest test branches with /tmp paths (test pollution)
+- Readme standard at 83% (only non-100% standard remaining)
 
 ## Architecture Notes
 - trigger branch owns error_detected events; ai_mail provides delivery only
 - extensions/, plugins/, json_templates/, json/ shim, persistence/ = repeatable skeleton (don't archive)
 - Dead code was inherited from DevPass migration, not accumulated in AIPass
+- .seedgo/bypass.json: email.py bypassed for introspection (inbox/sent/contacts need no args), json_utils/json_handler.py bypassed for json_structure (can't import itself)
 
 </details>
 
@@ -73,14 +76,14 @@
 
 </details>
 
-<details><summary><strong>@backup</strong> — Operational — seedgo 100% (all 23 standards) (2026-03-17 (session 9))</summary>
+<details><summary><strong>@backup</strong> — Operational — seedgo 100% (all 24 standards) (2026-03-17 (session 10))</summary>
 
 # @backup
 
 > Multi-mode backup — snapshot, versioned, Google Drive
 
-**State:** Operational — seedgo 100% (all 23 standards)
-**Last update:** 2026-03-17 (session 9)
+**State:** Operational — seedgo 100% (all 24 standards)
+**Last update:** 2026-03-17 (session 10)
 
 ## Working Commands
 - `drone @backup snapshot [--dry-run]` — full copy backup
@@ -92,6 +95,9 @@
 - 1590 files / 12.8 MB
 - 235 runtime JSON files filtered (correct behavior)
 
+## Recently Completed
+- Session 10: @seedgo dispatch — json_structure 4%→100%, introspection 60%→100%, readme freshness. Re-did session 9 work lost in git collision.
+
 ## Pending
 - @api dispatch: Migrate Google Drive auth to API branch (email c19c8d0f — parked)
 - integrations.py: locally archived to .archive but git state may have it back in backup.py imports — reconcile on next session
@@ -139,47 +145,64 @@
 
 </details>
 
-<details><summary><strong>@commons</strong> — Operational — Seedgo 99% (2026-03-17)</summary>
+<details><summary><strong>@commons</strong> — Operational — Seedgo 99% (2026-03-18)</summary>
 
 # @commons
 
 > Social network for branches — posts, rooms, artifacts
 
 **State:** Operational — Seedgo 99%
-**Last update:** 2026-03-17
+**Last update:** 2026-03-18
 
 ## Seedgo Compliance (99%)
 
 23/24 categories at 100%. Only Readme at 83% (minor tree check).
 
-| Category | Score |
-|----------|-------|
-| Architecture, Cli, Cli_Flags, Documentation | 100% |
-| Encapsulation, Error_Handling, Handlers, Imports | 100% |
-| Introspection, Json_Structure, Log_* (all 4) | 100% |
-| Meta, Modules, Naming, Permission_Flags | 100% |
-| Shebang, Stderr_Routing, Testing, Trigger | 100% |
-| Diagnostics | 100% |
-| **Readme** | **83%** |
+## Session 4 Summary (2026-03-17/18)
 
-## Session 3 Summary (2026-03-17)
-- DPLAN-0074: Full compliance & system verification
-- FPLAN-0076: Execution plan
-- Phase 0: Smoke test PASSED
-- Phase 1: Quick wins (stderr, readme date, dropbox, diagnostics)
-- Phase 2: Handler security guard (encapsulation 66%→100%)
-- Phase 3: Introspection gates (52%→100% with bypasses)
-- Phase 4: json_handler wired into 56 files (1%→100%)
-- Fixed log template bug (dict→list format)
-- Score: 91% → 99%
+### Prax Log Routing Fix
+- Stray `src/aipass/commons/` directory spotted by Patrick
+- Root cause: prax `detect_branch_from_path()` didn't handle commons' non-standard path
+- Dispatched to prax, who fixed 3 files. Verified: logs now write to `src/commons/logs/`
+
+### 5-Branch Social Stress Test
+- Tested with commons, drone, spawn, flow, memory
+- 40+ commands: posts, comments, votes, rooms, artifacts, explore, search
+- Identity detection via CWD: bulletproof across all 5 branches
+
+### 3 Integration Gaps Fixed (DPLAN-0018, FPLAN-0097)
+All same pattern: helper functions existed from Dev-Pass port but were never wired into production code.
+
+1. **FTS5 search** — `sync_post_to_fts()`/`sync_comment_to_fts()` never called from post_ops/comment_ops. Added calls + `backfill_fts_index()` (indexed 6 posts + 17 comments)
+2. **Visitor tracking** — No `room_visits` table. Added table to schema, `record_visit()` to space_ops, recording in `_cmd_enter()`. Updated `get_visitors_data()`.
+3. **Post/comment counts** — `increment_post_count()`/`increment_comment_count()` never called. Catchup nudge showed wrong "haven't posted" tip. Wired into post_ops/comment_ops.
+
+### Files Modified
+- `apps/handlers/posts/post_ops.py` — FTS sync + post count
+- `apps/handlers/comments/comment_ops.py` — FTS sync + comment count
+- `apps/handlers/search/search_queries.py` — backfill_fts_index()
+- `apps/handlers/rooms/space_ops.py` — record_visit() + updated visitors query
+- `apps/modules/space_module.py` — visit recording in enter
+- `apps/handlers/database/schema.sql` — room_visits table
+- `tools/backfill_fts.py` — one-time FTS backfill script
+
+## Session 5: Module Rename (2026-03-18)
+
+Removed `_module` suffix from all 21 module files per seedgo naming standard ("Path = Context, Name = Action"). DPLAN-0019 → FPLAN-0099 master plan.
+
+- 21 agents deployed in 2 batches (10 + 11), running in parallel
+- Each agent: renamed .py file, updated header + introspection string, renamed JSON tracking files
+- Orchestrator: updated commons.py static import, .seedgo/bypass.json (9 entries)
+- Old modules archived to `apps/modules/.archive/`
+- System verified: all commands working, seedgo 99%
 
 ## Known Issues
 - Readme 83% — likely missing dropbox/ in directory tree
 - Pyright warnings (unused logger imports) — pre-existing, cosmetic
 
 ## Architecture
-- 86 Python files, 21 modules, 19 handler domains
-- SQLite + WAL + FTS5, 16 tables
+- 86+ Python files, 21 modules, 19 handler domains
+- SQLite + WAL + FTS5, 17 tables (added room_visits)
 - Lives at src/commons/ (standalone, outside aipass namespace)
 - Imports own code as `from commons.apps.*`
 - Uses aipass services: prax (logging), cli (display)
@@ -223,61 +246,57 @@
 **Last update:** 2026-03-18
 
 ## Milestones
-- 35 sessions of system coordination
-- drone @git module — atomic PR workflow with lockfile (DPLAN-0008 complete)
-- STATUS board system (per-branch + central aggregation)
-- Prompt architecture (breadcrumbs pattern)
-- dev.local.md → STATUS.local.md consolidation
-- Dispatch+wake workflow (backup, flow, cli, prax all verified)
-- First PR review cycle — 8 PRs reviewed+merged in one session
+- 36 sessions of system coordination
+- drone @git pr v2 — commit-on-main workflow (no disappearing files)
+- Plan cleanup — 60+ FPLANs/DPLANs closed, templates updated with "close immediately" rule
+- Prax monitor fully operational — UNKNOWN→PRAX, inotify restored, polling fallback with actionable messaging
+- Global prompt refined — logging/debugging, .archive pattern, local-ahead-of-origin, console vs logs
+- System cleanup — 12 notepad items investigated and resolved (commons DB, memory central_writer, registry ghosts, .seed bypass, unknown_branch)
+- Bounce system confirmed working end-to-end
 
 ## Current Work
-- **DPLAN-0008 COMPLETE**: drone @git module built, tested, deployed. All 3 phases done. Global prompt updated, 13 branch deny lists deployed. Branches use `drone @git pr` only.
-- **Stale term cleanup**: Scanner tool built (`tools/dev_central_to_devpulse.py`). SOP written. 484 "dev central" hits across 155 files. 5 categories (A-E). Strategy: branch-by-branch, sub-agents handle A/B/E, plan C manually, leave D.
-- DPLAN-004: Dashboard pipeline — argparse bug still present (--all/--branch flags eaten by prax.py).
+- **Memory plan vectorization crisis**: Per-file subprocess cold-loads torch (~500MB) per plan file. 114 files × concurrent invocations = system crush. Wired into startup trigger so EVERY drone command fires it. Memory investigating with Dev-Pass comparison. Awaiting report.
+- **Dispatch UX ref sweep**: ai_mail + trigger updated. Prax, daemon, flow have emails in inbox (not yet woken).
+- **Local prompt gap**: 7 of 15 branches on 14-line stubs. Root cause of babysitting load vs Dev-Pass.
 - DPLAN-003: Phase 3 (drone aipass help) pending.
-- Ruff CI: 474 violations remaining. No cleanup until seedgo coverage higher.
 
 ## Known Issues
-- **prax dashboard CLI routing**: `drone @prax dashboard refresh --all` fails — argparse eats flags before module. Python API works fine.
-- **flow/DPLAN CWD default**: DPLANs always go to flow's dev_planning/ regardless of caller's CWD.
+- **CRITICAL: memory plan processing**: (1) per-file subprocess loads torch each time instead of batching, (2) no concurrency lock — multiple instances race same files, (3) wired into startup trigger — every drone command triggers heavy ML ops. Path fix in session 36 unmasked the bug (was silently finding 0 files before).
+- **prax dashboard CLI routing**: argparse eats flags before module.
 - **api**: `models` command not routed through drone
-- **ai_mail**: `get_current_user()` returns relative `mailbox_path` — causes doubled paths in reply
-- **drone**: versioned backup exceeds 30s subprocess timeout
-- **seedgo self-violation**: stderr_routing_check.py fails its own standard
-- **ruff CI**: 474 lint violations (331 unused imports, 87 unused vars, etc). Backlog — don't auto-fix until seedgo coverage is 100%.
-- **wake.py no --model flag**: dispatched branches use CLI default model. Can't control from our side.
+- **ruff CI**: 474 lint violations. Backlog — don't auto-fix until seedgo coverage is 100%.
+- **wake.py no --model flag**: dispatched branches use CLI default model.
 
 ## Todos
-- [ ] Close stale FPLANs (0017, 0021) once verified complete
+- [ ] Local prompt enrichment pass — 7 branches on stubs (drone, prax, cli, ai_mail, backup, api, spawn)
+- [ ] Devpulse wake protection — daemon blocks dispatch but manual wake has no blocklist. Need full protection like Dev-Pass DevCentral.
+- [x] DPLAN-0017: dispatch UX redesign — DELIVERED. dispatch/email/fresh all tested.
+- [ ] Handler guard too aggressive — blocks `python3 -c` even for same-branch agents. Guard should only block cross-branch .py file imports, not command-line execution. Seedgo owns the `__init__.py` template. All 14 branches affected.
 - [ ] DPLAN-003 Phase 3: `drone aipass help` module
-- [ ] Email prax about argparse bug (--all/--branch still broken despite PR #47)
-- [ ] Seedgo: single-file checklist command broken (`drone @seedgo checklist <file>` fails silently)
 - [ ] Ruff cleanup FPLAN (after seedgo coverage reaches 100%)
-- [ ] Add --model support to wake.py for dispatched branches
+- [ ] Add --model support to wake.py
 
 ## Recently Completed
-- Session 35: DPLAN-0008 complete — drone @git module built+tested+deployed, global prompt+deny lists, flow seedgo redo via drone @git pr. PRs #76-78 merged. (2026-03-18)
-- Session 34: CLAUDE.md inbox startup, culture doc refs, venv→symlinks, setup.sh symlink install, ghost input bug research (2026-03-17)
-- Session 33: Stale term scanner+SOP, dispatched ai_mail+trigger cleanup, PR #60 (2026-03-16)
-- Session 32: README update (init-anywhere), git workflow fix (return to main after PR — tested, works), merged PRs #54-56, cleaned 8 stale branches, new principle (2026-03-16)
-- Session 31: CLI front door (seedgo-compliant discovery + drone internal module), statusline git branch, Rich colors, DPLAN-0044, aipass init external test, 6 research agents, OpenClaw comparison, PR #51 (2026-03-15)
-- Session 29: Merged 8 PRs (#39-48), ruff investigation (612→474), dispatch reply breadcrumb, /prep broadened (2026-03-14)
-- Session 28: Global prompt governance, DPLAN/FPLAN, Git Workflow, How-to-Work, PR #41+#42 merged (2026-03-14)
-- Prax Phase 1+2: dashboard devpulse section removed, refresh pipeline 4 bugs fixed (2026-03-14)
-- CLI init_project.py migration: FPLAN-0041 done, 99% seedgo, all tests passed (2026-03-14)
-- FPLAN-0033 Phase 3 stderr migration — 10 branches, 48 files (2026-03-14)
-- DPLAN-003 Stage 1 — credential model complete (2026-03-14)
+- Session 36: PR v2 (commit-on-main). Global prompt (logging, .archive, git awareness). Notepad cleanup (12 items). Prax monitor fixed (UNKNOWN, inotify, introspection). Registry healed (6 ghosts removed). Memory central_writer path fixed. Commons DB path fixed. (2026-03-18)
+- Session 35: DPLAN-0008 complete — drone @git module. PRs #76-86. (2026-03-18)
+- Session 34: CLAUDE.md inbox startup, venv→symlinks. (2026-03-17)
+- Session 33: Stale term scanner+SOP, dispatched cleanup. (2026-03-16)
 
 ## Notepad
 > **SCRATCH SPACE — gets wiped at session start or topic change.**
 
-- Session 35 COMPLETE.
-- DPLAN-0008 fully done — all 3 phases. Module built, tested, prompt updated, deny lists deployed.
-- settings.local.json files are gitignored — deny lists are local enforcement only. Spawn template is source of truth for new clones.
-- memory_bank.central.json untracked in src/aipass/central/ — Patrick knows, needs to be moved.
-- Culture doc still pending Patrick review.
-- FPLAN-0094 should be closed (drone @git module complete).
+Session 37:
+- DPLAN-0017 DELIVERED + TESTED. dispatch/email/fresh all confirmed working on prax.
+- Dispatched ref updates to: ai_mail (done), trigger (done), flow (email landed, blocked by interactive), prax (killed mid-dispatch), daemon (killed mid-dispatch).
+- Memory vectorization incident: 5 concurrent drone commands each triggered startup→check_and_rollover→process_plans→per-file torch subprocess. System pegged at 100% CPU. Had to kill all processes. Root cause: session 36 path fix unmasked the bug — before fix, process_plans found 0 files and returned instantly.
+- Inbox hang was NOT ai_mail — confirmed. System was CPU-starved from memory's torch subprocesses.
+- Devpulse wake protection gap found: daemon has blocklist but manual wake doesn't.
+- Memory incident RESOLVED: batch mode + decoupled from startup + explicit command (PR #94, #95). 113 files processed in ~1 min at ~20% CPU.
+- Memory emailed trigger to review _run_memory_bank_check() in startup.py — still calls check_and_rollover() on every drone command.
+- Prax, daemon still have `send`→`email` ref update emails in their inboxes — low priority, will pick up next wake.
+- Flow has template update email + older pending items.
+- 100K+ trigger_core.log entries from the incident — prax handling the overflow.
+- PRs pending on GitHub: #89, #90, #94, #95, plus earlier flow/prax work.
 
 </details>
 
@@ -323,53 +342,72 @@
 
 </details>
 
-<details><summary><strong>@flow</strong> — In progress — seedgo compliance (FPLAN-0096) (2026-03-17)</summary>
+<details><summary><strong>@flow</strong> — In progress — health check (DPLAN-0020) (2026-03-18)</summary>
 
 # @flow
 
-> Plan lifecycle — FPLANs (building) + DPLANs (planning) via unified plugin architecture
+> Plan lifecycle — FPLANs, DPLANs, and any custom plan type via filesystem-driven template registry
 
-**State:** In progress — seedgo compliance (FPLAN-0096)
-**Last update:** 2026-03-17
+**State:** In progress — health check (DPLAN-0020)
+**Last update:** 2026-03-18
 
 ## Current Work
-- FPLAN-0096: Seedgo compliance — 40 tracked files modified
-  - json_handler wired into 33 handlers + 7 modules
-  - Introspection gates added to 7 modules
-  - Awaiting: zombie dplan cleanup, README update, final audit
+- DPLAN-0020: Flow health check — session 9 deep dive with Patrick
+  - Introspection/help: DONE (all 8 modules)
+  - Template rename + auto-discovery: DONE
+  - Template registry + auto-heal: DONE (delete dir → full cleanup)
+  - No-fallback errors: DONE
+  - --dry-run for close: DONE
+  - Registry cleanup: DONE (Docker artifacts, orphan FPLANs)
+  - Post-close pipeline: DONE (foreground archival, flags set atomically)
+  - Vector intake: DONE (process-plans + verify, visible in close output)
+  - Remaining: dashboard push fix, README update, devpulse template updates
+
+## Recently Completed (Session 9)
+- Foreground archival: moved from background subprocess to close_ops.py, fixing race condition
+- Registry flags: processed/cleanup_completed now set atomically on close
+- Vector pipeline: drone @memory process-plans + is_plan_vectorized() wired into close output
+- Template auto-heal: load_registry() prunes orphaned types + deletes plan registry JSON
+- --dry-run flag: drone @flow close --all --dry-run
+- Registry cleanup: removed 4 Docker-path entries, 7 orphan FPLANs archived
+- Silent fallback removed from mbank/process.py vector intake
+- Memory bug verified fixed (CWD), devpulse dispatch replied (--dry-run + orphan heal)
+- Closed DPLAN-0022, 0024, 0026
+- Cross-branch coordination: 3 email exchanges with @memory → path mismatch found + fixed
 
 ## Pending
-- Delete zombie `apps/handlers/dplan/` (restored from archive by agent, not tracked in git)
-- README update (do LAST with full system context, not just date bump)
-- Final seedgo audit to verify 100%
-- Reply to @memory bug report (email d2bde448: FPLAN create not registering)
+- 2 devpulse dispatches: FPLAN template syntax updates + list→list open docs
+- Dashboard push failing on close (non-critical, separate issue)
+- README update (template system, new commands, pipeline changes)
+- Final seedgo audit
 
 ## Active Plans
-- DPLAN-0016: Seedgo compliance design doc
-- FPLAN-0096: Seedgo compliance master plan
+- DPLAN-0020: Flow health check (near completion)
 
-## Open PR
-- None currently
-
-## Key Learnings (Session 6)
-- Work on main, only create branches at PR time via `drone @git pr`
-- Zombie files on disk inflate seedgo audit counts — 15 dplan files were already archived
-- Real tracked scope was 40 files, not 55
+## New Commands (Session 8-9)
+```
+drone @flow register <dir> <PREFIX>    # Register plan type
+drone @flow unregister <dir>           # Remove plan type
+drone @flow templates                  # List registered types
+drone @flow scan                       # Find unregistered directories
+drone @flow close --dry-run            # Preview what would close
+drone @flow close --all --dry-run      # Preview close-all
+```
 
 ## Known Issues
-- 64 orphaned plan files in flow/ root from before CWD fix (need cleanup after merge)
-- Default FPLAN template not always fast-deleted on close (cosmetic)
+- Dashboard push fails on every close ("Failed to push flow section to branch dashboard")
+- plan_type.json files still in templates/ dirs (orphaned — system ignores them)
 
 </details>
 
-<details><summary><strong>@memory</strong> — Operational — 100% seedgo compliant (2026-03-17)</summary>
+<details><summary><strong>@memory</strong> — Operational — 100% seedgo compliant (2026-03-18 (s16))</summary>
 
 # @memory
 
 > Vector memory archive — ChromaDB, sentence-transformers
 
 **State:** Operational — 100% seedgo compliant
-**Last update:** 2026-03-17
+**Last update:** 2026-03-18 (s16)
 
 ## Milestones
 - ChromaDB integration
@@ -377,96 +415,118 @@
 - Searchable vector archive
 - FPLAN-0026 complete (rollover E2E, search, plans archival)
 - **DPLAN-0073 COMPLETE** — 100% seedgo audit, all 24 standards passing
-- memory_json/ auto-populating with operation logs
+- **FPLAN-0128 COMPLETE** — verify command + config fix (PR #94)
+- **Session 16: CRITICAL FIX** — batch vectorization, decoupled from startup trigger (PR #95)
+- 113 plan files vectorized (2019 chunks) in ~1 min at ~20% CPU
 
 ## Current Work
 - No active DPLANs or FPLANs — all caught up
-- Completed: DPLAN-0073, DPLAN-0055, DPLAN-0045, FPLAN-0054, FPLAN-0062
+- Future: wire up watchdog watcher via `drone @memory watch` to replace startup trigger dependency
 
 ## Known Issues
 - `search` fails without torch/sentence-transformers installed
-- DPLANs not vectorized on close (flow-side issue)
-- Flow plan registry broken — `drone @flow create` makes file but doesn't register. Close fails. Emailed @flow.
-- Old memory bank had ~4,180 vectors. Current has 127. Full port pending.
+- Search returns semantically similar plans, not exact ID matches — use `verify` for ID lookup
+- ~~Plan vectorization system overload~~ FIXED in PR #95 (batch mode + decoupled from startup)
+- ~~Plan vectorization path mismatch~~ FIXED in PR #95 (correct path: src/aipass/backup/processed_plans)
+- ~~Flow plan registry broken~~ FIXED in PR #71
+- Old memory bank had ~4,180 vectors. Current has 2,200+ (207 rollover + 2019 plans). Growing.
+
+## New Commands
+- `drone @memory verify FPLAN-XXXX` — check if plan is vectorized in chroma
+- `drone @memory process-plans` — explicit batch vectorization (NOT auto-triggered)
+
+## Architecture Notes
+- Dual vector storage: local (`{branch}/.chroma/`) + global (`memory/.chroma/`). Local fails gracefully, global is critical (restores backup on failure).
+- Plan vectorization decoupled from startup trigger. `_check_plans()` = count only. Processing via explicit command.
+- Watchdog infrastructure exists in memory_watcher.py but not exposed via drone yet.
+- Startup trigger chain: prax logger → trigger.fire('startup') → check_and_rollover(). Keep this lightweight — NEVER put heavy processing here.
 
 ## Notepad
-- Session 12: 100% seedgo audit achieved. All commands tested and working.
-- memory_json/ now has 8 log files auto-created from rollover check/status commands
-- spawn has template_owners.json (empty) — memory should populate it as authoritative source for .trinity files
+- Session 16: INCIDENT — vectorization overloaded system. 228 subprocess calls × 5 concurrent drone commands. Fixed with batch mode + decoupling. PR #95. Full report sent to devpulse.
+- Session 15: FPLAN-0128 — verify command + config fix. PR #94.
+- Session 14: Processed 4 @flow emails. Path mismatch root cause found.
+- Emailed trigger about startup.py _run_memory_bank_check() weight — needs review
 - handlers/__init__.py guard blocks cross-branch imports — always expose via modules layer
-- Post-rollover chain wired: Trigger.fire → central_writer → dashboard_push → pool_processor
-- Memory at 28/25 key_learnings — rollover will trigger on next addition
 
 </details>
 
-<details><summary><strong>@prax</strong> — Operational (2026-03-17)</summary>
+<details><summary><strong>@prax</strong> — Operational (2026-03-18)</summary>
 
 # @prax
 
 > Logging, monitoring, dashboard infrastructure
 
 **State:** Operational
-**Last update:** 2026-03-17
+**Last update:** 2026-03-18
 
 ## Milestones
 - System-wide logging via `from aipass.prax import logger`
-- Monitor command (passive + interactive)
+- Monitor: polling fallback, caller attribution, agent tracking (Phases 1-3)
 - Dashboard infrastructure
-- Two-tier logging simplification (PR #59, merged)
-- Seedgo audit: 93% → 98% (session 13, PR #65 pending)
+- Seedgo audit: 100% (25/25 categories)
 
 ## Current Work
-- PR #65 pending merge (json_structure + stderr + dead code cleanup)
-- 48 active files (9 dead handlers archived)
+- No active plans — all closed
+- 4 active commands: monitor, status, log-audit, dashboard
+- 5 dead modules archived (init, shutdown, run, terminal, discover)
 
-## Seedgo Audit: 98%
-- 23 of 24 categories at 100%
-- Log_Structure: false positives — seedgo PR #63 fixes the checker
-- After both PRs merge (#63 + #65) → expect 100%
+## Seedgo Audit: 100%
+- 25 of 25 categories at 100%
 
-## Pending Cross-Branch
-- @seedgo PR #63 (log_structure checker fix) — confirmed fixed, awaiting merge
+## Recently Completed (Session 17)
+- Introspection standard: all modules compliant
+- Monitor Phase 1: PollingObserver fallback, degradation notices, clean startup
+- Monitor Phase 2: Caller attribution via drone [CALLER:BRANCH] markers
+- Monitor Phase 3: Agent intelligence (tool use, thinking, responses)
+- Archived 5 dead modules, stripped verbose flag, cleaned help
+- Found + reported seedgo silent checker bug — fixed same session
+- Final 3 introspection violations fixed (status.py, agent_status.py, logger.py)
+- Cleaned stale logs/.archive/
 
 ## Known Issues
-- None blocking
+- System inotify watches near-exhausted. Monitor uses polling fallback — functional but slower.
+- Phase 4 (interactive filtering) deferred — not needed now
 
 </details>
 
-<details><summary><strong>@seedgo</strong> — Operational — **100% compliance** (2026-03-17 (session 21))</summary>
+<details><summary><strong>@seedgo</strong> — Operational — **13/14 branches at 100%** (but new introspection checks will lower some scores) (2026-03-18 (session 22))</summary>
 
 # @seedgo
 
-> Standards enforcement, 24-standard audit pack + hook ownership
+> Standards enforcement, 24-standard audit pack + hook ownership + branch dispatch
 
-**State:** Operational — **100% compliance**
-**Last update:** 2026-03-17 (session 21)
+**State:** Operational — **13/14 branches at 100%** (but new introspection checks will lower some scores)
+**Last update:** 2026-03-18 (session 22)
 
-## Session 21 Summary
+## Session 22 Summary
 
-1. **log_structure fix** — Removed hierarchical placement post-check from branch_audit.py. Was penalizing prax for having logs. Prax now 100%.
-2. **pre_edit_gate.py v1.1** — Branch-scoped. Cross-branch type errors no longer block edits.
-3. **Hook audit + DPLAN-0012** — Inventoried all 12 hooks. Found duplicate PostToolUse, stale docs, bridge architecture discussion. Patrick decided: seedgo owns hooks.
-4. **Portability architecture** — Bridge pattern (project_bridge.sh → prompt_inject.sh) is the pip install solution. NOT dead scripts. Auto-fix is AIPass-specific, stays in project hooks.
-5. **branch_audit.py type fixes** — Path | None, Dict[str, Any]. Phantom revert occurred and was redone.
-6. **Emailed Memory** — json_handler clarification for DPLAN-0073.
+Dispatch wake — processed 3 emails.
+
+1. **Introspection checker improved** (Flow dispatch) — Two new checks:
+   - **Content references** — flags `print_introspection()`/`print_help()` that reference `python3` instead of `drone @` commands
+   - **Module help interception** — flags modules where `handle_command()` doesn't intercept `--help`
+2. **Standard content fixed** — `introspection_content.py` examples updated from `python3` to `drone @`
+3. **Devpulse HOLD** — Introspection standard stays as-is. Standards changes only from devpulse/Patrick.
+4. **Prax rejected** — Told to use subcommands instead of removing introspection guards.
 
 ## Active Plans
 
-- **DPLAN-0012** — Hook Management. PARKED. Architecture mapped, not executing yet. Patrick has other work in progress.
-- **DPLAN-0075** — Auto-fix Execution. Two-hook system working in production. Monitor for edge cases.
+- **DPLAN-0012** — Hook Management. PARKED. Architecture mapped, not executing yet.
+- **DPLAN-0075** — Auto-fix Execution. Working in production.
 
 ## TODO
 
-- **DPLAN-0012 execution** — When Patrick is ready. Phase 1: duplicate fix, Phase 2: cross-branch gate, Phase 3: docs
-- **seedgo_verify modernization** — old verify module in .archive/, needs rebuild
-- **log_structure fix review** — Patrick had mixed feelings on PR #63 fix
+- **ai_mail readme** — 99% → 100% (only remaining non-100% branch)
+- **System-wide introspection fixes** — New checks will flag python3 refs + missing --help interception across branches
+- **DPLAN-0012 execution** — When Patrick is ready
+- **seedgo_verify modernization** — old verify module in .archive/
+- **Update branch prompt** — Add dispatch patterns, hook ownership responsibilities
 
 ## Known Issues
-- Duplicate PostToolUse hooks — auto_fix fires twice per edit (PARKED, don't touch)
+- Duplicate PostToolUse hooks — auto_fix fires twice per edit (PARKED)
 - Hardcoded `/home/patrick/` paths in global settings hooks
-- Bridge scripts (project_bridge.sh, prompt_inject.sh) superseded but NOT dead — portability solution
-- Phantom file revert on branch_audit.py — cause unknown, watching
-- readme_check.py command list check stops at ### sub-headings
+- Bridge scripts superseded but needed for portability
+- commons at src/commons/ not src/aipass/commons/ — path assumption trap
 
 </details>
 
diff --git a/src/aipass/api/apps/handlers/auth/env.py b/src/aipass/api/apps/handlers/auth/env.py
index 256eb039..d64d2d74 100644
--- a/src/aipass/api/apps/handlers/auth/env.py
+++ b/src/aipass/api/apps/handlers/auth/env.py
@@ -152,17 +152,17 @@ def create_env_template(provider: str = "openrouter", target_path: Optional[Path
 
     Args:
         provider: API provider name (default: 'openrouter')
-        target_path: Optional custom path (defaults to api/.env)
+        target_path: Optional custom path (defaults to ~/.secrets/aipass/.env)
 
     Returns:
         bool: True if template created successfully, False otherwise
 
     Example:
         >>> if create_env_template('openrouter'):
-        ...     print("Template created at <api_root>/.env")
+        ...     print("Template created at ~/.secrets/aipass/.env")
     """
-    # Default to api/.env
-    env_path = target_path or (API_ROOT / ".env")
+    # Default to ~/.secrets/aipass/.env (cross-platform standard)
+    env_path = target_path or (Path.home() / ".secrets" / "aipass" / ".env")
 
     # Don't overwrite existing file
     if env_path.exists():
diff --git a/src/aipass/api/apps/handlers/auth/keys.py b/src/aipass/api/apps/handlers/auth/keys.py
index 5d5288c6..2bfab7e3 100644
--- a/src/aipass/api/apps/handlers/auth/keys.py
+++ b/src/aipass/api/apps/handlers/auth/keys.py
@@ -268,6 +268,53 @@ def get_validation_rules(provider: str) -> Dict[str, Any]:
 # KEY FORMAT CHECKING
 # ==============================================
 
+def diagnose_key(provider: str = "openrouter") -> str:
+    """
+    Diagnose why get_api_key() returned None.
+
+    Checks all sources for a raw key (skipping validation) and explains
+    exactly why it failed — missing entirely, wrong prefix, too short, etc.
+
+    Args:
+        provider: Provider name (default: 'openrouter')
+
+    Returns:
+        str: Human-readable explanation of the key issue
+
+    Example:
+        >>> if not get_api_key('openrouter'):
+        ...     print(diagnose_key('openrouter'))
+    """
+    # Check all sources for raw key (without validation)
+    key = get_key_from_config(provider)
+    source = "config"
+
+    if not key:
+        key = get_key_from_env(provider)
+        source = "env"
+
+    if not key:
+        env_var = f"{provider.upper()}_API_KEY"
+        key = read_env_file(env_var)
+        source = "dotenv"
+
+    if not key:
+        return "No API key found in any source (config, environment, .env file)"
+
+    # Key exists but failed validation — explain why
+    key = key.strip()
+    rules = get_validation_rules(provider)
+
+    if "prefix" in rules and not key.startswith(rules["prefix"]):
+        actual_prefix = key[:len(rules["prefix"])] if len(key) >= len(rules["prefix"]) else key[:6]
+        return f"Key found ({source}) but invalid — expected prefix '{rules['prefix']}', got '{actual_prefix}...'"
+
+    if "min_length" in rules and len(key) < rules["min_length"]:
+        return f"Key found ({source}) but too short — {len(key)} chars, need {rules['min_length']}+"
+
+    return f"Key found ({source}) but failed validation"
+
+
 def check_key_format(key: str) -> Dict[str, Any]:
     """
     Analyze key format and return details.
diff --git a/src/aipass/api/apps/handlers/openrouter/client.py b/src/aipass/api/apps/handlers/openrouter/client.py
index c82c5d90..e467d837 100644
--- a/src/aipass/api/apps/handlers/openrouter/client.py
+++ b/src/aipass/api/apps/handlers/openrouter/client.py
@@ -56,6 +56,7 @@ except ImportError:
 # Handler imports
 from aipass.api.apps.handlers.auth.keys import get_api_key
 from aipass.api.apps.handlers.openrouter.caller import get_caller_info
+from aipass.api.apps.handlers.openrouter.provision import ensure_caller_config
 from aipass.api.apps.handlers.usage.tracking import track_usage
 
 # JSON handler
@@ -315,6 +316,12 @@ def get_response(prompt: str, caller: Optional[str] = None, model: Optional[str]
             # logger.warning("Could not detect caller - using 'unknown'")
             caller = "unknown"
 
+    # Step 1b: Ensure caller has config (auto-provision if missing)
+    try:
+        ensure_caller_config(caller)
+    except Exception:
+        pass  # Provisioning is best-effort — don't block the API call
+
     # Step 2: Require model from caller - no defaults
     if not model:
         # logger.error("No model specified - caller must provide model from their branch config")
diff --git a/src/aipass/api/apps/modules/api_key.py b/src/aipass/api/apps/modules/api_key.py
index d8ace3e7..03aa3b4b 100644
--- a/src/aipass/api/apps/modules/api_key.py
+++ b/src/aipass/api/apps/modules/api_key.py
@@ -153,9 +153,15 @@ def init_env():
     header("Initialize API Configuration")
     console.print()
 
+    env_path = Path.home() / ".secrets" / "aipass" / ".env"
+
+    if env_path.exists():
+        success(f"Environment file already exists at {env_path}")
+        return
+
     # Create .env template via handler
     if env.create_env_template():
-        success("Environment template created")
+        success(f"Environment template created at {env_path}")
     else:
         error("Failed to create environment template")
 
diff --git a/src/aipass/api/apps/modules/openrouter_client.py b/src/aipass/api/apps/modules/openrouter_client.py
index 23334ae0..5a2343a8 100644
--- a/src/aipass/api/apps/modules/openrouter_client.py
+++ b/src/aipass/api/apps/modules/openrouter_client.py
@@ -138,7 +138,7 @@ def handle_command(command: str, args: List[str]) -> bool:
             test_connection()
             return True
         if command == "models":
-            list_models()
+            list_models(args)
             return True
         if command == "status":
             check_status()
@@ -164,17 +164,23 @@ def test_connection():
     header("Test OpenRouter Connection")
     console.print()
 
+    console.print("[dim]Testing connection...[/dim]")
+
     # Get API key via handler
     api_key = keys.get_api_key("openrouter")
 
     if not api_key:
-        error("No API key configured")
+        diagnosis = keys.diagnose_key("openrouter")
+        error(diagnosis)
         return
 
-    warning("Testing connection...")
+    # Real API ping — hit /models endpoint
+    model_list = models.fetch_models_from_api(api_key)
 
-    # TODO: Call handler to test connection when implemented
-    success("Connection test successful")
+    if model_list:
+        success(f"Connection successful — {len(model_list)} models available")
+    else:
+        error("Connection failed — could not reach OpenRouter API")
 
 
 def make_call(args: List[str]):
@@ -187,31 +193,67 @@ def make_call(args: List[str]):
     warning("API call workflow - TODO")
 
 
-def list_models():
+def list_models(args: List[str] | None = None):
     """Orchestrate list models workflow"""
     header("Available Models")
     console.print()
 
+    show_all = args and "--all" in args
+
     # Get API key via handler
     api_key = keys.get_api_key("openrouter")
 
     if not api_key:
-        error("No API key configured")
+        diagnosis = keys.diagnose_key("openrouter")
+        error(diagnosis)
         return
 
-    warning("Fetching available models...")
+    console.print("[dim]Fetching available models...[/dim]")
 
     # Call handler to fetch models
     model_list = models.fetch_models_from_api(api_key)
 
-    if model_list:
-        success(f"Found {len(model_list)} models")
-        for model in model_list[:10]:
-            console.print(f"  - {model}")
-        if len(model_list) > 10:
-            console.print(f"  ... and {len(model_list) - 10} more")
-    else:
+    if not model_list:
         error("Failed to fetch models")
+        return
+
+    success(f"Found {len(model_list)} models")
+    console.print()
+
+    # Format as table
+    display_count = len(model_list) if show_all else min(10, len(model_list))
+
+    console.print(f"  {'Model':<50} {'Context':>10} {'$/prompt':>10} {'$/compl':>10}")
+    console.print(f"  {'─' * 50} {'─' * 10} {'─' * 10} {'─' * 10}")
+
+    for model_data in model_list[:display_count]:
+        model_id = model_data.get("id", "unknown")
+        context = model_data.get("context_length", 0)
+        pricing = model_data.get("pricing", {})
+        prompt_cost = pricing.get("prompt", "0")
+        completion_cost = pricing.get("completion", "0")
+
+        # Format context length
+        if context >= 1_000_000:
+            ctx_str = f"{context // 1_000_000}M"
+        elif context >= 1_000:
+            ctx_str = f"{context // 1_000}k"
+        else:
+            ctx_str = str(context)
+
+        # Format pricing
+        if str(prompt_cost) == "0" and str(completion_cost) == "0":
+            p_str = "free"
+            c_str = "free"
+        else:
+            p_str = f"${prompt_cost}"
+            c_str = f"${completion_cost}"
+
+        console.print(f"  {model_id:<50} {ctx_str:>10} {p_str:>10} {c_str:>10}")
+
+    if not show_all and len(model_list) > 10:
+        console.print()
+        console.print(f"  [dim]Showing 10 of {len(model_list)} — use --all for full list[/dim]")
 
 
 def check_status():
@@ -219,8 +261,32 @@ def check_status():
     header("OpenRouter Client Status")
     console.print()
 
-    # TODO: Call handlers for status when implemented
-    warning("Client status - TODO")
+    # Key status
+    api_key = keys.get_api_key("openrouter")
+
+    if api_key:
+        masked = api_key[:8] + "..." + api_key[-4:]
+        console.print(f"  [cyan]Key configured:[/cyan]  [green]yes[/green]")
+        console.print(f"  [cyan]Key:[/cyan]             {masked}")
+    else:
+        console.print(f"  [cyan]Key configured:[/cyan]  [red]no[/red]")
+        diagnosis = keys.diagnose_key("openrouter")
+        console.print(f"  [cyan]Reason:[/cyan]          {diagnosis}")
+
+    console.print(f"  [cyan]Provider:[/cyan]        OpenRouter")
+    console.print(f"  [cyan]Base URL:[/cyan]        https://openrouter.ai/api/v1")
+
+    # OpenAI SDK availability
+    try:
+        import openai  # noqa: F401
+        console.print(f"  [cyan]OpenAI SDK:[/cyan]     [green]available[/green]")
+    except ImportError:
+        console.print(f"  [cyan]OpenAI SDK:[/cyan]     [red]missing[/red]")
+
+    # Client cache stats
+    cache_stats = client.get_cache_stats()
+    console.print(f"  [cyan]Cached clients:[/cyan] {cache_stats['cached_clients']}/{cache_stats['max_cache_size']}")
+    console.print()
 
 
 # =============================================
diff --git a/src/aipass/devpulse/docs/README.md b/src/aipass/devpulse/docs/README.md
new file mode 100644
index 00000000..dad7824b
--- /dev/null
+++ b/src/aipass/devpulse/docs/README.md
@@ -0,0 +1,22 @@
+# DPLAN-003 Working Directory
+
+Research, mapping, and planning files for "AIPass as Operating System."
+
+Parent plan: `AIPass/DPLAN-003_aipass_as_operating_system_2026-03-13.md`
+
+## Scope
+
+**In scope:** `src/aipass/` branches only (drone, seedgo, prax, cli, flow, ai_mail, api, trigger, spawn, devpulse, backup, daemon, memory).
+
+**Out of scope:** `src/commons/`, `src/skills/` — these are separate and not part of this refactor.
+
+## Files
+
+| File | Purpose | Status |
+|------|---------|--------|
+| `registry_discovery_map.md` | Every find_registry() call in src/aipass/ | Done |
+| `portability_audit.md` | Full investigation results (session 24) | Done |
+| `credential_model.md` | Registry credential design (UUID now, macaroons later) | Stage 1 Done |
+| `registry_refactor_plan.md` | Shared find_registry() design, migration steps | Pending |
+| `aipass_init_spec.md` | CLI owns init — `drone @cli aipass init` (entry point later) | Pending |
+| `aipass_help_spec.md` | CLI owns help — `drone @cli aipass help` (entry point later) | Pending |
diff --git a/src/aipass/devpulse/docs/ai_mail_comms_upgrade.md b/src/aipass/devpulse/docs/ai_mail_comms_upgrade.md
new file mode 100644
index 00000000..91d9d108
--- /dev/null
+++ b/src/aipass/devpulse/docs/ai_mail_comms_upgrade.md
@@ -0,0 +1,217 @@
+# AI_MAIL COMMS UPGRADE — Hardening Plan
+
+**Goal:** Make ai_mail bulletproof, then use as the model for other branches.
+**Started:** 2026-03-10 | **Status:** In Progress
+**Tracking:** Updated each session. Read this first on context resume.
+
+---
+
+## Current State (Session 10)
+
+**Seedgo audit: 100%** — all 23 standards pass (42 files)
+**Automated tests: 36** — test_send_identity.py v1.2.0 (3 audit rounds, 15 agents, 0 false positives)
+**Production fixes deployed:** 3 cross-platform crashers fixed, test suite fully isolated
+**Phase 2 (Error Handling):** 27 silent `except` blocks across 13 files now log with `logger.warning()`
+
+### Test Suite Evolution
+- v1.0.0: 31 tests — 7 false positives found by 3 audit agents
+- v1.1.0: 32 tests — 8 suspects found by 5 audit agents (live registry, weak contracts)
+- v1.2.0: 36 tests — final audit found 2 minor issues, fixed. **0 live-data dependencies.**
+
+### Production Fixes (Session 9)
+- `inbox_lock.py` — `import fcntl` guarded for Windows (msvcrt fallback)
+- `ai_mail.py` — `signal.SIGPIPE` guarded with `hasattr` check
+- `notify.py` — `/usr/bin/python3` replaced with `shutil.which("python3")`
+
+---
+
+## Architecture Map
+
+```
+Entry: apps/ai_mail.py
+  +-- Module: apps/modules/email.py (orchestrator, v3.0.0)
+       |-- handle_send() -> send_args.py (parse) -> send.py (execute) -> delivery.py (inbox write)
+       |-- handle_inbox() -> inbox_resolve.py -> inbox_ops.py -> format.py
+       |-- handle_view() -> inbox_ops.py
+       |-- handle_reply() -> reply.py -> delivery.py
+       |-- handle_close() -> close_ops.py
+       |-- handle_sent() -> format.py
+       +-- handle_contacts() -> registry/
+
+  Module: apps/modules/dispatch.py (dispatch orchestrator)
+       |-- daemon.py (auto-dispatch loop)
+       |-- wake.py (manual branch wake)
+       |-- dispatch_monitor.py (agent lifecycle)
+       |-- status.py (dispatch status display)
+       +-- pending_work.py (pending dispatch queue)
+
+  Identity: apps/handlers/users/
+       |-- branch_detection.py (detect_branch_from_pwd — THE critical path)
+       |-- user.py (get_current_user, get_branch_by_email)
+       +-- load.py, config_generator.py
+
+  Cross-branch: drone/apps/handlers/router_handler.py
+       +-- detect_caller_branch_name() -> sets AIPASS_CALLER_BRANCH env var
+```
+
+**Identity detection chain (9 stages):**
+1. drone CLI entry
+2. router resolves @branch to path
+3. router_handler detects caller via CWD (+ AIPASS_BRANCH_NAME fallback)
+4. executor merges caller_env into subprocess
+5. ai_mail subprocess starts with env vars
+6. branch_detection reads AIPASS_CALLER_BRANCH
+7. send_args builds headers
+8. delivery writes to recipient inbox.json
+9. notify.py fires desktop notification
+
+---
+
+## Systemic Issues Found (Session 9 Deep Audit)
+
+15 agents across 3 rounds audited tests + full codebase. Key findings:
+
+### Silent Failures (24 instances) — VIOLATES "fail to errors" rule
+`except Exception: return None` in 10+ places with zero logging:
+- `branch_detection.py` — 3 functions (entire identity chain)
+- `delivery.py` — get_all_branches(), summary, notification
+- `create.py` — load_email_file()
+- `user.py` — get_user_by_email(), get_all_users()
+- `daemon.py` — _read_json() (silent) vs wake.py (logs) — inconsistent
+- `inbox_ops.py` — migration persist has literal `pass`
+
+### Dead/Redundant Code (~20% of codebase, ~1,728 lines)
+- 5 fully unused files: validate.py, errors.py, data_ops.py, config_generator.py, pending_work.py
+- `_find_repo_root()` copy-pasted in 9 files
+- `get_all_branches()` implemented twice with different email behavior (correctness bug)
+- `lock_utils.py` exists but unused — wake.py and daemon.py reimplement locking
+- wake.py/daemon.py share 6+ duplicated functions
+
+### Cross-Platform Breakers
+- [FIXED] `import fcntl` unconditional — crashes Windows
+- [FIXED] `/usr/bin/python3` hardcoded — breaks macOS/Windows
+- [FIXED] `signal.SIGPIPE` — crashes Windows startup
+- [OPEN] `pgrep` ungated in wake.py/daemon.py
+- [OPEN] `email.py` parents[2] fragile — should use _find_repo_root()
+
+---
+
+## Hardening Phases
+
+### Phase 1: Test Suite (Critical Path)
+**Priority: HIGH** | **Status: IN PROGRESS**
+
+- [x] `test_send_identity.py` — 36 tests, 3 audit rounds, 0 false positives
+- [ ] `test_delivery.py` — round-trip send/receive
+- [ ] `test_send_args.py` — argument parsing (all flags, interactive, error cases)
+- [ ] `test_inbox_ops.py` — inbox operations (view, close, reply, close all)
+- [ ] `test_dispatch_monitor.py` — dispatch lifecycle
+- [ ] `test_notify.py` — notification delivery
+
+### Phase 2: Error Handling Overhaul
+**Priority: HIGH** | **Status: IN PROGRESS**
+Silent failures are why the system feels "fragile."
+
+- [x] Add `logger.warning()` to every bare `except Exception: return None` — **13 files, 27 instances fixed**
+- [ ] Distinguish "not found" from "error reading" in return types
+- [ ] Fix inconsistent error conventions (None vs tuple vs dict vs raise)
+- [ ] Fix collision detection dead code in delivery.py get_all_branches()
+
+### Phase 3: Code Consolidation
+**Priority: MEDIUM** — Reduce duplication, single source of truth.
+
+- [ ] Extract shared `_find_repo_root()` into commons or shared utility
+- [ ] Consolidate `get_all_branches()` — one implementation, prefers explicit email
+- [ ] Consolidate lock acquisition — use lock_utils.py, delete reimplementations
+- [ ] Deduplicate wake.py/daemon.py shared functions (_read_json, _set_session_name, etc.)
+- [ ] Archive 5 dead files (validate.py, errors.py, data_ops.py, config_generator.py, pending_work.py)
+
+### Phase 4: Identity Consolidation
+**Priority: MEDIUM** — Reduce 12 detection mechanisms to 1 canonical resolver.
+
+- [ ] Define canonical `resolve_branch_identity()` function
+- [ ] Priority chain: AIPASS_CALLER_BRANCH > AIPASS_BRANCH_NAME > CWD passport walk > --from
+- [ ] Single file, single function, single source of truth
+- [ ] All callers delegate to it
+
+### Phase 5: Cross-Platform Hardening
+**Priority: MEDIUM** — Public repo must work on all platforms.
+
+- [ ] Guard `pgrep` usage in wake.py/daemon.py
+- [ ] Replace `email.py` parents[2] with _find_repo_root()
+- [ ] Guard `start_new_session=True` for Windows
+- [ ] Add encoding='utf-8' to os.fdopen in wake.py
+
+### Phase 6: Standards for Communication
+**Priority: LOW** — Make patterns enforceable system-wide.
+
+- [ ] `communication` standard — envelope format, required fields, identity rules
+- [ ] `identity` standard — env var hierarchy, passport requirements
+- [ ] `dispatch` standard — env isolation, lock management, bounce handling
+
+### Phase 7: Documentation & System Prompt
+**Priority: LOW** — Update branch local prompt.
+
+- [ ] Write proper `.aipass/aipass_local_prompt.md`
+- [ ] Document key commands, architecture, critical files
+
+---
+
+## Decision Log
+
+| Date | Decision | Rationale |
+|------|----------|-----------|
+| 2026-03-08 | AIPASS_BRANCH_NAME env var over CWD detection | CWD unreliable when agents navigate. Env vars persist. |
+| 2026-03-08 | dbus direct over notify-send | Portal mode strips persistence hints. dbus bypasses confinement. |
+| 2026-03-08 | Unique app_name per source | GNOME collapses same desktop-entry into one notification. |
+| 2026-03-08 | Fail loud, no CWD fallback for identity | Silent wrong sender worse than visible error. |
+| 2026-03-10 | --from flag for explicit sender override | Plumbing existed, just needed CLI wiring. |
+| 2026-03-10 | Tests first in hardening plan | Every bug from sessions 4-7 would have been caught by tests. |
+| 2026-03-10 | 3 audit rounds on test suite | Each round found issues the previous missed. Diminishing returns by round 3. |
+| 2026-03-10 | Error handling overhaul before code consolidation | Silent failures cause more user pain than code duplication. |
+
+---
+
+## Session Notes
+
+### Session 10 (2026-03-10)
+- **Phase 2: Error Handling Overhaul — logger.warning() sweep complete**
+- 13 files modified, 27 silent `except Exception` blocks now log with `logger.warning()`
+- Files fixed (by priority):
+  - `branch_detection.py` (3) — identity chain, most critical
+  - `user.py` (2) — user lookup
+  - `delivery.py` (6) — get_all_branches, migrate, callback, summary, notification, private branch
+  - `create.py` (2) — purge, load_email_file
+  - `inbox_ops.py` (1) — migration persist
+  - `format.py` (1) — alias lookup
+  - `reply.py` (1) — get_email_by_id
+  - `close_ops.py` (3) — dashboard, central, purge post-ops
+  - `send.py` (2) — central update after send/broadcast
+  - `dashboard_sync.py` (1) — push_dashboard_update
+  - `inbox_cleanup.py` (4) — migrate, dashboard, central, purge
+  - `error_handler.py` (1) — error notification delivery
+  - `error_dispatch.py` (2) — dashboard, central post-ops
+- All 13 files pass `py_compile` syntax check
+- Remaining silent blocks (acceptable): inbox_lock.py finally-block, json_handler.py (utility),
+  daemon/wake/dispatch_monitor (already log with logger.info), config_generator/data_ops (unused files)
+- Next: Phase 2 remaining items (return type distinctions, error conventions, collision dead code)
+
+### Session 9 (2026-03-10, continued)
+- Wrote test_send_identity.py v1.0.0 (31 tests)
+- Audit round 1: 3 agents found 7 false positives -> rewrote to v1.1.0
+- Audit round 2: 5 agents found 8 issues (live registry, weak contracts) -> v1.2.0 (36 tests)
+- Audit round 3: 5 agents (final test + full system sweep)
+  - Tests: 2 minor issues found and fixed (unpatched registry, missing mailbox_path assert)
+  - Error handling: 24 silent failure instances across 10+ files
+  - Code quality: ~1,728 dead/redundant lines (20% of codebase), 5 unused files
+  - Cross-platform: 3 CRITICAL (fcntl, SIGPIPE, /usr/bin/python3) — ALL FIXED
+  - Paths: parents[N] usage audited, mostly correct, 1 fragile case in email.py
+- Fixed 3 cross-platform crashers in production code
+- Updated COMMS_UPGRADE.md with full findings and revised phases
+- Next: Phase 1 continues (more test files) or Phase 2 (error handling overhaul)
+
+### Session 8 (2026-03-10)
+- Patrick initiated hardening project
+- Self-audit: 100% on all 22 seedgo standards
+- Mapped full architecture (55 Python files, 3 modules, 8 handler domains)
+- Created this tracking document
diff --git a/src/aipass/devpulse/docs/credential_model.md b/src/aipass/devpulse/docs/credential_model.md
new file mode 100644
index 00000000..53757f31
--- /dev/null
+++ b/src/aipass/devpulse/docs/credential_model.md
@@ -0,0 +1,157 @@
+# DPLAN-003: Registry Credential Model
+
+## The Idea
+
+Registries get a unique token. Passports carry that token. Access is identity-based, not filesystem-based. No walk-up needed — your credential proves which registry is yours.
+
+## Why
+
+Current system finds registries by walking up directories. Works for one project, breaks with multiple. If two AIPass projects exist on one machine, a citizen launched from the wrong directory finds the wrong registry. Credentials solve this — your passport carries proof of membership.
+
+## Prior Art (Research)
+
+| System | Pattern | Fit |
+|--------|---------|-----|
+| **Macaroons** (Google Research) | Token IS the credential. Delegatable with caveats. Offline verification. DeepMind validated for AI agent delegation (2026). | Highest |
+| **Vault Namespaces** | Project = namespace. Token scoped to namespace. Mini-registry per project. | High |
+| **AWS STS / Token Vending** | Agent presents project ID, gets scoped credential. Credential itself is the boundary. | High |
+| **K8s Namespace + ServiceAccount** | Token carries project scope as claim. RBAC composable. | High |
+| **SPIFFE/SPIRE** | Process-level attestation without static secrets. | Medium |
+| **direnv** | Auto-set env vars on directory entry. Zero-friction UX. | UX pattern |
+
+Full research: agent output from session 25.
+
+## Design: Two Stages
+
+### Stage 1: UUID Match (Manual, Now)
+
+Simple. Prove the concept works before adding crypto.
+
+**Registry gets an ID:**
+```json
+{
+  "metadata": {
+    "id": "a1b2c3d4-...",
+    "name": "AIPASS",
+    "version": "1.0.0",
+    "last_updated": "2026-03-13",
+    "total_branches": 15
+  },
+  "branches": [...]
+}
+```
+
+**Passports get the matching ID:**
+```json
+{
+  "citizenship": {
+    "registered": true,
+    "registry_id": "a1b2c3d4-...",
+    "registry_name": "AIPASS",
+    "citizen_number": 7
+  }
+}
+```
+
+**Lookup flow:**
+1. Walk up from CWD, find `*_REGISTRY.json`
+2. Read its `metadata.id`
+3. Check citizen's `citizenship.registry_id` matches
+4. If mismatch → error ("citizen belongs to registry X, found registry Y")
+5. If match → proceed
+
+**Spawn changes:**
+- `aipass init` (or manual setup) generates the UUID for the registry
+- Spawn reads registry UUID and injects into new passports via `{{REGISTRY_ID}}` placeholder
+- Existing 15 branches get the UUID added to their passports (one-time migration)
+
+### Stage 2: Macaroon Tokens (Future, When Cross-Project Needed)
+
+Upgrade path when we need delegation and cross-project access.
+
+**Root token:** Created at `aipass init`. HMAC-based. Stored at `~/.secrets/aipass/projects/<uuid>.token`
+
+**Citizen token:** Attenuated copy in passport. Can prove membership but can't mint new citizens.
+
+**Agent token:** Further attenuated. Carries:
+- Registry ID (which project)
+- Scope (full access — agents do the real work in AIPass)
+- Expiry (session-scoped, dies when agent dies)
+- Issuer (which citizen spawned this agent)
+
+Agents are NOT read-only. They're the builders — they write code, run tests, modify files. The token proves they belong to a project, it doesn't restrict what they do within it. Scope restrictions would be role-based (e.g., "can't modify other branches' files") not capability-based.
+
+**Verification:** Local HMAC check. No daemon needed for basic validation. Daemon is optional enhancement for audit logging and revocation.
+
+**direnv integration:** Entering a project directory auto-sets `AIPASS_PROJECT_TOKEN` in env. Agents inherit it.
+
+## What Changes (Stage 1)
+
+| Component | Change |
+|-----------|--------|
+| `AIPASS_REGISTRY.json` | Add `metadata.id` (UUID) |
+| Passport template | Add `citizenship.registry_id` placeholder |
+| Spawn `build_replacements_dict()` | Read registry UUID, add `{{REGISTRY_ID}}` |
+| Spawn `add_to_registry()` | No change (branch entries stay the same) |
+| Drone `find_registry()` | Optional: verify passport.registry_id matches found registry |
+| All 15 passports | One-time: add `registry_id` field |
+
+## What Does NOT Change
+
+- Registry filename stays `*_REGISTRY.json`
+- Walk-up discovery still works (credential is verification layer on top, not replacement)
+- Branch structure unchanged
+- No daemon needed
+- No new dependencies
+
+## Resolved Questions
+
+1. **Registry ID location** → `metadata.id` — it's the project's identity, not the citizen's.
+2. **UUID4 vs hash** → UUID4 (random). Simple, guaranteed unique, no inputs needed.
+3. **Verification mode** → Hard error on mismatch. Fail loudly — that's the AIPass way.
+4. **Where does init live?** → Temporary: `src/aipass/devpulse/apps/init_project.py`. Future: CLI branch.
+
+## Bugs, Quirks, and Findings
+
+Discovered during testing. Reference for future work.
+
+### init_project.py (10 edge cases tested)
+- **Spaces in dir name** → registry filename gets spaces (`MY COOL PROJECT_REGISTRY.json`). Fixed: `_sanitize_name()` replaces non-alphanumeric with `_`.
+- **Root path `/`** → `Path("/").name` is empty string, creates `_REGISTRY.json`. Fixed: validation rejects empty name.
+- **Permission errors** → raw traceback instead of clean message. Fixed: `main()` catches `OSError`.
+- **Passport overwrite on re-init** → if registry deleted but `.trinity/` survives, re-init would silently overwrite passport with new UUID. Fixed: passport guarded with `exists()` check.
+- **Double init** → correctly blocked by `FileExistsError` on registry file.
+- **Deep nested paths** → `mkdir(parents=True)` handles correctly.
+- **UUID uniqueness** → 5 runs, 5 unique UUIDs. No collisions.
+- **JSON validity** → all generated files parse clean.
+
+### Drone isolation (6 scenarios tested)
+- **Sibling projects** → PASS. Two projects in same parent dir, fully isolated.
+- **Nested project** → PASS. Inner registry wins over outer. No bleed-through.
+- **Deep subdir walk-up** → PASS. Finds nearest ancestor registry correctly.
+- **Cross-project contamination** → PASS. Branches are registry-scoped; modules are global.
+- **Empty directory (no registry)** → CONCERN. Drone silently falls back to AIPass source registry via `__file__` walk-up. Any dir on this machine without a registry sees production branches. This is by design in `find_registry()` but will be dangerous with multi-project. Credential verification would catch this — citizen's registry_id won't match the fallback registry.
+- **`drone @ai_mail` from mock project** → correctly returns "Branch not found in registry" (empty project has no branches).
+
+### Registry/passport gitignore
+- `AIPASS_REGISTRY.json` and all `.trinity/passport.json` files are gitignored. UUID migration is local-only. This is correct for now — credentials are machine-specific, not repo state. But means `aipass init` must run on every clone/install. Future: consider whether UUID should be in-repo or machine-local.
+
+### AI mail after migration
+- Send/receive works fine after UUID migration. AI mail's 4 internal `find_registry()` copies still hardcode `AIPASS_REGISTRY.json` — functional for now since that filename exists, but won't find `*_REGISTRY.json` in other projects.
+
+### Drone built-in modules vs branches
+- `@drone` and `@seedgo` are hardcoded as "modules" in `module_registry.py`, always visible everywhere. Other branches (`@ai_mail`, `@spawn`, etc.) are registry-scoped. This distinction matters: modules are global services, branches are project citizens.
+
+## Decision Log
+
+- **2026-03-14:** Patrick proposed credential-based registry access. Agents do the real work — tokens prove membership, not restrict capability. Manual first, `aipass init` later.
+- **2026-03-14:** Research confirmed macaroons as best-fit pattern (Google Research + DeepMind 2026 validation). Stage 1 = UUID match, Stage 2 = macaroon upgrade.
+- **2026-03-14:** Scope limited to `src/aipass/` branches only. Commons and skills excluded.
+- **2026-03-14:** All 4 open questions resolved. `init_project.py` built and tested — creates registry with UUID, passport with matching registry_id, .trinity/, .aipass/, AIPASS.md. Tested with temp directory — drone isolation confirmed (only built-in modules visible, not AIPass branches).
+- **2026-03-14:** UUID migration executed (FPLAN-0030). Registry + 13 passports updated. Registry and passports are gitignored — UUID is machine-local, not repo state.
+- **2026-03-14:** Drone verification added (`_verify_registry_credential` in load_registry). Drone dispatched and fixed error propagation — new `RegistryMismatchError` class separates "not found" (fallback OK) from "mismatch" (hard error). Spawn templates updated with `{{REGISTRY_ID}}` placeholder.
+- **2026-03-14:** Stage 1 complete. Credential model is end-to-end: init creates credentialed registries, drone verifies on load, spawn injects into new passports, mismatch = hard error. Remaining: drone CLI stderr surfacing (DPLAN-0031), `aipass init` CLI command (future).
+- **2026-03-14:** FPLAN-0032 executed — CLI stderr standardization Phase 1+2 done. CLI owns err_console, error()/warning() go to stderr. Drone imports from CLI instead of creating its own Console(stderr=True). Credential mismatch errors now properly surface to terminal via stderr. The stderr issue that blocked error visibility during DPLAN-003 testing is resolved.
+- **2026-03-14:** CLI confirmed as owner of credential model + project commands. CLI owns: `aipass init` (create project), `aipass help` (project help), display API (error/warning/fatal/err_console). Drone owns verification (registry_handler). Spawn owns injection (passport templates). Access via `drone @cli aipass init/help` for now; `aipass init` entry point is future icing.
+- **2026-03-14:** Seedgo `stderr_routing` standard created (24th standard). Checker bug found — was scanning all files but not displaying module/handler results. Seedgo fixed display + aggregation. Agent-verified across 3 branches. FPLAN-0033 parked for Phase 3 migration (343 error prints, 10 branches).
+- **2026-03-14:** FPLAN-0033 Phase 3 executed autonomously (Patrick away). 15+ agents across 3 waves migrated 10 branches. 48 files, 666 insertions. System stderr avg 49%→69%. Post-migration verification found 63% of remaining seedgo violations are false positives ([yellow]Label:[/yellow] section headers flagged as warnings). Seedgo dispatched for checker refinement. 93 real violations remain (48 in commons which was out of scope).
diff --git a/src/aipass/devpulse/docs/portability_audit.md b/src/aipass/devpulse/docs/portability_audit.md
new file mode 100644
index 00000000..090f00cc
--- /dev/null
+++ b/src/aipass/devpulse/docs/portability_audit.md
@@ -0,0 +1,31 @@
+# Portability Audit — Session 24 Results
+
+## Summary
+
+| Tool | Registry Discovery | CWD-Aware | Portable | Hardcoded |
+|------|-------------------|-----------|----------|-----------|
+| Drone | Walk-up + env var | No (uses registry) | Yes | Registry filename |
+| Spawn | Walk-up + env var | No (uses registry) | Partial | Template location |
+| Prax | Walk-up (no env) | No (sys logs at repo) | Partial | System logs dir |
+| AI_Mail | Walk-up (no env) | No (inbox per branch) | Yes | Inbox location |
+| Flow | Walk-up (no env) | Yes (plan creation) | Hybrid | Plan registry |
+
+## Key Findings
+
+- All tools use walk-up strategy to find `AIPASS_REGISTRY.json`
+- Registry-relative path resolution already works (move registry + dirs = works)
+- `AIPASS_REGISTRY` env var supported by drone and spawn
+- System logs hardcoded to `{repo_root}/system_logs/`
+- Spawn templates hardcoded to `{spawn_package}/templates/`
+- Walk-up doesn't stop at project boundaries — finds nearest registry up the tree
+
+## The Core Fix
+
+Change `find_registry()` to:
+1. Walk up from CWD looking for `*_REGISTRY.json` (glob, not hardcoded name)
+2. Stop at first match — that's the project boundary
+3. If none found, return error ("No AIPass project. Run `aipass init`")
+
+## Source
+
+Full investigation transcript: background agent session 24, 42 tool calls across drone/spawn/ai_mail/flow/prax.
diff --git a/src/aipass/devpulse/docs/registry_discovery_map.md b/src/aipass/devpulse/docs/registry_discovery_map.md
new file mode 100644
index 00000000..17583c4c
--- /dev/null
+++ b/src/aipass/devpulse/docs/registry_discovery_map.md
@@ -0,0 +1,46 @@
+# Registry Discovery Map — FPLAN-0029 Phase 1
+
+## Primary Implementations (5)
+
+| File | Function | Strategy |
+|------|----------|----------|
+| `src/aipass/spawn/apps/handlers/registry.py:34-79` | `find_registry(start_path)` | Env var → project markers (.git/pyproject.toml) → walk-up → cwd fallback |
+| `src/aipass/drone/apps/handlers/registry_handler.py:36-61` | `find_registry()` | Walk-up from __file__ → walk-up from cwd → parents[4] fallback |
+| `src/aipass/drone/apps/handlers/registry_handler.py:64-81` | `get_registry_path()` | Global override → env var → find_registry() |
+| `src/commons/apps/handlers/database/db.py:343-378` | `_find_branch_registry()` | Env var AIPASS_ROOT → walk-up (10 limit) → ~/.aipass/ fallback |
+| `src/commons/apps/handlers/identity/identity_ops.py:33-67` | `_find_branch_registry_path()` | Same as db.py |
+
+## Local Reimplementations (23+)
+
+All do walk-up from `__file__` looking for `AIPASS_REGISTRY.json`:
+
+### ai_mail (4 files)
+- `apps/handlers/users/branch_detection.py:26-38`
+- `apps/handlers/registry/read.py:31-41`
+- `apps/handlers/email/format.py:21-31`
+- `apps/handlers/dispatch/daemon.py:36-55`
+
+### memory (3 files)
+- `apps/handlers/dashboard_push.py:44-54`
+- `apps/handlers/monitor/detector.py:36-83`
+- `apps/handlers/monitor/memory_watcher.py:363-385`
+
+### prax (2 files)
+- `apps/handlers/registry/reader.py:24-39`
+- `apps/handlers/dashboard/agent_status_writer.py:33-44`
+
+### seedgo (3 files)
+- `apps/handlers/audit/discovery.py:54-62`
+- `apps/handlers/diagnostics/discovery.py:21-29`
+- `apps/handlers/readme/readme_ops.py:29-37`
+
+### + 11 more across other branches
+
+## Hardcoded String Count
+100+ references to `"AIPASS_REGISTRY.json"` across codebase.
+
+## Phase 1 Plan
+1. Create `commons.registry.find_registry()` — one function, glob for `*_REGISTRY.json`
+2. All 23+ modules import from commons
+3. Drone/spawn wrappers call commons internally
+4. Test: temp dir with TEST_REGISTRY.json → drone systems finds it
diff --git a/src/aipass/devpulse/docs/system_health_report.md b/src/aipass/devpulse/docs/system_health_report.md
new file mode 100644
index 00000000..5764d1da
--- /dev/null
+++ b/src/aipass/devpulse/docs/system_health_report.md
@@ -0,0 +1,525 @@
+# AIPass System Health Report
+Generated: 2026-03-19 22:54
+
+---
+
+## 1. Dead Code (14 unused files)
+
+
+Dead Code Scanner
+Scanning 15 branches
+
+@ai_mail (3 modules, 31 handlers)
+  x handlers/monitoring/errors.py -- 0 references
+  > 33/34 files referenced
+
+@api (4 modules, 14 handlers)
+  x handlers/openrouter/provision.py -- 0 references
+  > 17/18 files referenced
+
+@backup (4 modules, 24 handlers)
+  x handlers/diff/vscode_integration.py -- 0 references
+  > 27/28 files referenced
+
+@cli (3 modules, 2 handlers)
+  > All 5 files referenced
+
+@commons (22 modules, 35 handlers)
+  > All 57 files referenced
+
+@daemon (6 modules, 8 handlers)
+  > All 14 files referenced
+
+@devpulse -- no apps/ directory
+@drone (9 modules, 16 handlers)
+  > All 25 files referenced
+
+@flow (8 modules, 35 handlers)
+  x handlers/plan/file_ops.py -- 0 references
+  x handlers/plan/update_registry.py -- 0 references
+  > 41/43 files referenced
+
+@memory (5 modules, 29 handlers)
+  x handlers/learnings/manager.py -- 0 references
+  x handlers/schema/normalize.py -- 0 references
+  x handlers/search/vector_search.py -- 0 references
+  > 31/34 files referenced
+
+@prax (6 modules, 36 handlers)
+  > All 42 files referenced
+
+@seedgo (5 modules, 66 handlers)
+  x handlers/config/aipass_bypass.py -- 0 references
+  x handlers/config/aipass_ignore.py -- 0 references
+  x handlers/diagnostics/python_diognostics.py -- 0 references
+  x handlers/diagnostics/typscript_diognostics.py -- 0 references
+  x handlers/file/file_handler.py -- 0 references
+  x handlers/mock_standard_1/bypass_config/bypass.config.py -- 0 references
+  > 65/71 files referenced
+
+@skills (5 modules, 8 handlers)
+  > All 13 files referenced
+
+@spawn (6 modules, 15 handlers)
+  > All 21 files referenced
+
+@trigger (5 modules, 17 handlers)
+  > All 22 files referenced
+
+TOTAL: 14 unused across 14 branches
+
+---
+
+## 2. Local Prompts (5 stubs need enrichment)
+
+
+Local Prompt Status
+==================================================
+
+RICH (50+ lines, 4+ sections):
+  v daemon            56 lines   4 sections
+  v devpulse          79 lines   8 sections
+  v flow              67 lines   6 sections
+  v seedgo            56 lines   4 sections
+  v skills            58 lines   5 sections
+  v trigger           53 lines   7 sections
+
+BASIC (15-49 lines):
+  ~ api               15 lines   2 sections
+  ~ commons           48 lines   6 sections
+  ~ memory            37 lines   5 sections
+  ~ spawn             20 lines   4 sections
+
+STUB (<15 lines):
+  x ai_mail           14 lines   1 section 
+  x backup            14 lines   1 section 
+  x cli               14 lines   1 section 
+  x drone             14 lines   1 section 
+  x prax              14 lines   1 section 
+
+==================================================
+SUMMARY: 6 rich, 4 basic, 5 stub
+
+==================================================
+Section Breakdown
+==================================================
+
+@ai_mail (14 lines, STUB)
+  v Status
+    470 bytes
+
+@api (15 lines, BASIC)
+  v Identity
+  v Key Breadcrumbs
+    1192 bytes
+
+@backup (14 lines, STUB)
+  v Status
+    499 bytes
+
+@cli (14 lines, STUB)
+  v Status
+    499 bytes
+
+@commons (48 lines, BASIC)
+  v Commands
+  v Architecture
+  v Integration Points
+  v Critical Files
+  v Role
+  v Key Details
+    2328 bytes
+
+@daemon (56 lines, RICH)
+  v Commands
+  v Memory & Tracking
+  v Apps Layout
+  v Known Issues
+    2788 bytes
+
+@devpulse (79 lines, RICH)
+  v Identity
+  v How You Work
+  v Dispatch Table
+  v Commands
+  v Branches
+  v Working Habits
+  v Autonomous Monitoring
+  v Memory & Tracking
+  v Has dispatch table (@branch refs)
+    4710 bytes
+
+@drone (14 lines, STUB)
+  v Status
+    503 bytes
+
+@flow (67 lines, RICH)
+  v Commands
+  v Architecture
+  v Integration Points
+  v Conventions
+  v Critical Files
+  v Plan Type System
+    3461 bytes
+
+@memory (37 lines, BASIC)
+  v Identity
+  v Commands
+  v Architecture
+  v Memory & Tracking
+  v Known Issues
+    1591 bytes
+
+@prax (14 lines, STUB)
+  v Status
+    501 bytes
+
+@seedgo (56 lines, RICH)
+  v Commands
+  v Apps Layout (extra layer vs standard branch)
+  v How I Work — Standards Reasoning
+  v Quick Reference
+    3166 bytes
+
+@skills (58 lines, RICH)
+  v Commands
+  v Memory & Tracking
+  v Apps Layout
+  v Search Paths (first match wins)
+  v Three Skill Tiers
+    2706 bytes
+
+@spawn (20 lines, BASIC)
+  v Commands
+  v Architecture
+  v Role
+  v Principles
+    745 bytes
+
+@trigger (53 lines, RICH)
+  v Dispatch Table
+  v Commands
+  v Architecture
+  v Integration Points
+  v Critical Files
+  v Role
+  v Rules
+    2610 bytes
+
+
+---
+
+## 3. Commands (173 discovered)
+
+
+@ai_mail (14 commands)
+  - close
+  - contacts
+  - dispatch
+  - email
+  - inbox
+  - ping
+  - read
+  - registry
+  - reply
+  - send
+  - sent
+  - status
+  - thresholds
+  - view
+
+@api (12 commands)
+  - call
+  - cleanup
+  - google
+  - init
+  - models
+  - reauth
+  - session
+  - stats
+  - status
+  - test
+  - track
+  - validate
+
+@backup (1 command)
+  - reauth
+
+@cli (5 commands)
+  - aipass
+  - demo
+  - display
+  - show
+  - templates
+
+@commons (51 commands)
+  - activity
+  - artifacts
+  - capsule
+  - capsules
+  - catchup
+  - collab
+  - comment
+  - craft
+  - database
+  - decorate
+  - delete
+  - digest
+  - drop
+  - enter
+  - event
+  - explore
+  - feed
+  - find
+  - gift
+  - inspect
+  - leaderboard
+  - leaderboards
+  - log
+  - look
+  - mint
+  - mute
+  - open
+  - pin
+  - pinned
+  - post
+  - preferences
+  - profile
+  - prompt
+  - react
+  - reactions
+  - room
+  - search
+  - secrets
+  - sign
+  - thread
+  - track
+  - trade
+  - trending
+  - unpin
+  - unreact
+  - visitors
+  - vote
+  - watch
+  - welcome
+  - who
+  - whoami
+
+@daemon (5 commands)
+  - actions
+  - activity
+  - activity_report
+  - schedule
+  - update
+
+@drone (24 commands)
+  - activate
+  - add
+  - branches
+  - check
+  - exists
+  - info
+  - list
+  - load
+  - lock
+  - lookup
+  - path
+  - pr
+  - remove
+  - reset
+  - resolve
+  - route
+  - route_all
+  - scan
+  - set
+  - status
+  - sync
+  - system
+  - systems
+  - unlock
+
+@flow (11 commands)
+  - aggregate
+  - close
+  - create
+  - list
+  - post_close
+  - register
+  - registry
+  - restore
+  - scan
+  - templates
+  - unregister
+
+@memory (13 commands)
+  - analyze
+  - bootstrap
+  - check
+  - demo
+  - extract
+  - fragments
+  - rollover
+  - search
+  - status
+  - symbolic
+  - templates
+  - verify
+  - watch
+
+@prax (4 commands)
+  - dashboard
+  - log-audit
+  - monitor
+  - status
+
+@seedgo (8 commands)
+  - audit
+  - checklist
+  - diagnostics
+  - diagnostics_audit
+  - readme
+  - readme_update
+  - standards_audit
+  - standards_query
+
+@skills (6 commands)
+  - create
+  - discover
+  - info
+  - list
+  - run
+  - validate
+
+@spawn (4 commands)
+  - create
+  - delete
+  - passport
+  - update
+
+@trigger (15 commands)
+  - branch_log_events
+  - core
+  - errors
+  - fire
+  - list
+  - log_events
+  - medic
+  - mute
+  - off
+  - on
+  - reset
+  - start
+  - status
+  - stop
+  - unmute
+
+DISCOVERED: 173 commands across 14 branches
+
+---
+
+## 4. Test Coverage (26% module coverage)
+
+
+@ai_mail (49 tests, 2 files)
+  tests/test_send_identity.py              -- 36 tests  -> email, users
+  tests/test_user_paths.py                 -- 13 tests  -> users
+  UNTESTED: branch_ping, central_writer, dispatch, json, json_utils, monitoring, notify, registry
+
+@api (0 tests, 0 files)
+  (no test files found)
+
+@backup (0 tests, 1 file)
+  tests/test_pattern_scan.py               --  0 tests  -> config, operations
+  UNTESTED: backup_core, config, diff, google_drive_sync, integrations, json, models, operations, reauth_drive, reporting, utils
+
+@cli (0 tests, 0 files)
+  (no test files found)
+
+@commons (82 tests, 2 files)
+  tests/test_commons.py                    -- 72 tests  -> curation, database, notifications, profiles, search, welcome
+  tests/test_lifecycle.py                  -- 10 tests  -> database, search
+  UNTESTED: activity, artifact, artifacts, capsule, catchup, central, comment, comments, commons_identity, dashboard, digest, engagement, explore, feed, identity, json, leaderboard, notification, post, posts, profile, reaction, room, rooms, social, space, trade
+
+@daemon (31 tests, 1 file)
+  tests/test_actions_registry.py           -- 31 tests  -> actions
+  UNTESTED: activity_report, json, monitoring, schedule, scheduler_ops, update, wakeup_ops
+
+@devpulse (0 tests, 0 files)
+  (no test files found)
+
+@drone (335 tests, 9 files)
+  tests/test_activation.py                 -- 38 tests  -> command_registry, executor
+  tests/test_commands.py                   -- 35 tests  -> command_registry, commands
+  tests/test_discovery.py                  -- 40 tests  -> discovery, discovery_handler, exceptions, module_registry_handler
+  tests/test_executor.py                   -- 26 tests  -> exceptions, executor
+  tests/test_git_module.py                 -- 46 tests  -> git, git_module, module_registry_handler
+  tests/test_registry_handler.py           -- 33 tests  -> exceptions, registry_handler
+  tests/test_resolver.py                   -- 52 tests  -> exceptions, registry_handler, resolver
+  tests/test_router.py                     -- 37 tests  -> exceptions, executor, router, router_handler
+  tests/test_scan.py                       -- 28 tests  -> scan, scanning
+  UNTESTED: config, json, module_registry, registry
+
+@flow (0 tests, 0 files)
+  (no test files found)
+
+@memory (0 tests, 0 files)
+  (no test files found)
+
+@prax (0 tests, 0 files)
+  (no test files found)
+
+@seedgo (0 tests, 0 files)
+  (no test files found)
+
+@skills (123 tests, 8 files)
+  tests/test_cli_routing.py                -- 20 tests  -> ?
+  tests/test_discovery.py                  -- 32 tests  -> discovery_handler
+  tests/test_lifecycle.py                  -- 12 tests  -> creator, discovery, loader_handler, runner, template
+  tests/test_loader.py                     --  7 tests  -> loader
+  tests/test_registry.py                   -- 14 tests  -> registry
+  tests/test_runner.py                     -- 13 tests  -> runner
+  tests/test_runner_handler.py             -- 14 tests  -> runner_handler
+  tests/test_validator.py                  -- 11 tests  -> validator
+  UNTESTED: creator_handler, json
+
+@spawn (113 tests, 5 files)
+  tests/test_citizen_classes.py            -- 28 tests  -> class_registry, core, passport_ops, update, update_ops
+  tests/test_handlers.py                   -- 36 tests  -> change_detection, json_ops, meta_ops, reconcile
+  tests/test_lifecycle.py                  -- 22 tests  -> delete, delete_ops, sync_registry, sync_registry_ops, sync_templates, sync_templates_ops
+  tests/test_spawn.py                      -- 13 tests  -> metadata, placeholders, registry
+  tests/test_update.py                     -- 14 tests  -> meta_ops, update, update_ops
+  UNTESTED: file_ops, json, passport
+
+@trigger (0 tests, 0 files)
+  (no test files found)
+
+
+Test Coverage Report
+═══════════════════════════════════════════════════════
+
+TESTED:
+  ✓ drone            335 tests    9 files   15/19 modules covered (79%)
+  ✓ skills           123 tests    8 files   10/12 modules covered (83%)
+  ✓ spawn            113 tests    5 files   18/21 modules covered (86%)
+
+PARTIAL:
+  ◐ ai_mail           49 tests    2 files   2/10 modules covered (20%)
+  ◐ commons           82 tests    2 files   6/33 modules covered (18%)
+  ◐ daemon            31 tests    1 file    1/8 modules covered (12%)
+
+NO TESTS:
+  ✗ api                0 tests    0 files   0/10 modules covered (0%)
+  ✗ backup             0 tests    1 file    0/11 modules covered (0%)
+  ✗ cli                0 tests    0 files   0/5 modules covered (0%)
+  ✗ devpulse           0 tests    0 files   0/0 modules covered (0%)
+  ✗ flow               0 tests    0 files   0/16 modules covered (0%)
+  ✗ memory             0 tests    0 files   0/16 modules covered (0%)
+  ✗ prax               0 tests    0 files   0/14 modules covered (0%)
+  ✗ seedgo             0 tests    0 files   0/14 modules covered (0%)
+  ✗ trigger            0 tests    0 files   0/12 modules covered (0%)
+
+═══════════════════════════════════════════════════════
+SUMMARY: 733 tests across 15 branches
+         3 branches tested, 3 partial, 9 untested
+         Coverage: 52/201 modules (26%)
+
diff --git a/src/aipass/devpulse/docs/testing_standards.md b/src/aipass/devpulse/docs/testing_standards.md
new file mode 100644
index 00000000..f807717a
--- /dev/null
+++ b/src/aipass/devpulse/docs/testing_standards.md
@@ -0,0 +1,337 @@
+# Testing Standards
+**Status:** Draft v1 - Current manual process documented
+**Date:** 2025-11-12
+
+---
+
+## Current State: Manual Testing (Effective for Rapid Iteration)
+
+**Reality:** AIPass is a custom framework in active evolution. System changes weekly. Building extensive test infrastructure now = updating tests constantly instead of building features.
+
+**Current approach:** Manual testing with JSON/log verification works effectively for rapid iteration phase.
+
+**Future:** pytest framework expansion once branches stabilize (pytest is already configured and operational in some branches).
+
+---
+
+## The 90% Build Process
+
+**How we build and test:**
+
+### 1. Planning Phase (Upfront Investment)
+- Issue plan through Flow (master plan or default plan depending on scope)
+- Spend time getting structure right before coding
+- Define what we're building and how it fits together
+
+### 2. Build to 90% (AI-Led, Internal Verification)
+- AI builds structure and implementation
+- **Internal verification tests as we go:**
+  - Does the module turn on?
+  - Do commands work?
+  - Basic functionality confirmed?
+- Human mostly observes, provides input if things go astray
+- Focus on getting structure and pieces in place
+
+**Linux advantage:** Same environment (AI and human both on Linux) = outputs match = internal tests are reliable. Windows had issues with this, Linux doesn't.
+
+### 3. 90% Threshold (Human Review Phase)
+- Human gets actively involved
+- Reviews structure and implementation
+- Asks questions about concerns
+- Performs feature tests (what module is supposed to do)
+- Identifies bugs and missing error handling
+
+### 4. Debug Cycle (Error Handling First, Then Fixes)
+
+**Critical pattern: Fix error handling BEFORE fixing bugs**
+
+**Example scenario:**
+```
+Bug: API call not working, but console says "Success"
+No error output, but logs show API never executed
+Output is lying
+
+Process:
+1. Fix error handling FIRST - make errors tell the truth
+2. THEN fix the actual bug (why API isn't executing)
+3. See clean pass with honest outputs
+4. Move to next feature
+```
+
+**Why this order:**
+- Can't debug effectively if errors lie
+- Truth in outputs = faster debugging
+- Honest error messages = system teaches itself what's wrong
+
+### 5. Iterate Until Acceptable
+- Test different features
+- Continue debug cycle (handle errors → fix bugs → verify)
+- Reach acceptable standard (basic or advanced depending on module)
+- Move on
+
+---
+
+## Why Manual Testing Works Right Now
+
+**Advantages:**
+- **Fast iteration** - No test maintenance overhead
+- **Flexible** - System can change tomorrow without breaking test suite
+- **JSON/log infrastructure** - Acts as verification layer
+  - Check config.json → see settings
+  - Check data.json → see state
+  - Check log.json → see operations history
+  - Check Prax logs → detailed debugging
+- **Linux environment match** - AI tests = human tests (same outputs)
+- **Effective debugging** - Error-first approach catches issues early
+
+**Tradeoffs:**
+- Manual effort required at 90% stage
+- No automated regression testing (yet)
+- Relies on human review for final verification
+
+**Acceptable because:** System is evolving rapidly. Better to build fast and test manually than build slow and maintain brittle tests.
+
+---
+
+## Future: pytest Framework
+
+**Infrastructure in place:**
+- `pytest.ini` at repo root - Configuration for test discovery and markers
+- `tests/` - Root test directory with conftest.py
+- Branch-specific test directories at `src/aipass/{branch}/tests/` (API, Prax, CLI, etc.)
+
+**Already operational in some branches:**
+- API branch: 4 test files (test_api_system.py, test_openrouter_key.py, etc.)
+- Prax branch: test_log_rotation.py validates rotation behavior
+- Seedgo branch: test_cli_errors.py demonstrates error handling patterns
+
+**When to expand testing:**
+- Once modules and branches stabilize
+- When system changes slow down (monthly, not weekly)
+- When maintenance cost < value of automation
+
+**Why pytest:**
+- Standard Python testing framework
+- Already configured with pytest.ini
+- Infrastructure exists, ready for expansion
+- Some branches already have working tests
+
+**Current selective approach:**
+- Test critical/stable components (API, Prax log rotation)
+- Skip testing rapidly changing features
+- Manual testing for experimental work
+- Automated tests where they add value without maintenance burden
+
+---
+
+## Error Handling Philosophy
+
+**Errors must tell the truth** - foundational to testing effectiveness
+
+**Good error handling:**
+```python
+try:
+    result = api_call()
+    if not result:
+        logger.error("API call failed - no response")
+        return {"success": False, "error": "API returned no data"}
+except Exception as e:
+    logger.error(f"API call exception: {e}", exc_info=True)
+    return {"success": False, "error": str(e)}
+```
+
+**Bad error handling:**
+```python
+try:
+    result = api_call()
+    return {"success": True}  # LIES - didn't check if result valid
+except:
+    pass  # Silent failure - no truth
+```
+
+**Testing relies on honest errors:**
+- If errors lie, testing is impossible
+- Fix error handling first = testing becomes possible
+- Then fix bugs with confident verification
+
+---
+
+## JSON/Log Infrastructure as Testing Layer
+
+**Three-JSON pattern supports testing:**
+
+### Config Verification
+```bash
+# Check if settings are correct
+cat module_name_config.json
+# See API keys, limits, feature toggles
+```
+
+### State Verification
+```bash
+# Check current state
+cat module_name_data.json
+# See metrics, counts, current status
+```
+
+### Operations Verification
+```bash
+# Check what actually happened
+cat module_name_log.json
+# See recent operations and results
+```
+
+### Detailed Debugging
+```bash
+# Prax provides file-based logging, not real-time watching
+# Check system logs directory for detailed output
+ls -la system_logs/
+# Read specific module logs (Prax manages this directory location)
+cat system_logs/module_name.log
+```
+
+**This infrastructure = verification layer without formal tests**
+
+---
+
+## Testing Workflow Example
+
+**Building a new branch creation module:**
+
+1. **Plan** - Define structure, features, workflow (Flow plan)
+
+2. **Build to 90%** - AI implements:
+   - Module structure
+   - Handler functions
+   - Config/data/log JSONs
+   - Internal verification: "Does `create_branch test_branch` work?"
+
+3. **Review at 90%** - Human tests:
+   - Create branch with various names
+   - Check if files copied correctly
+   - Verify registry updated
+   - Try edge cases (existing branch, invalid name)
+
+4. **Find bug** - Branch created but registry not updated
+   - **First:** Check error handling - is error logged? Is return value honest?
+   - Add error handling if missing
+   - **Then:** Fix bug - why isn't registry updating?
+   - Verify with clean pass
+
+5. **Iterate** - Test more features:
+   - Template copying
+   - Placeholder replacement
+   - Memory file handling
+   - Backup on conflicts
+
+6. **Acceptable** - All major features work, errors are honest, ready to use
+
+---
+
+## Current Testing Checklist
+
+**For any new module/feature:**
+
+- [ ] Does it turn on without errors?
+- [ ] Do basic commands work?
+- [ ] Are errors handled and logged?
+- [ ] Do outputs tell the truth?
+- [ ] Check config.json - settings correct?
+- [ ] Check data.json - state tracking working?
+- [ ] Check log.json - operations recorded?
+- [ ] Test edge cases (invalid input, missing files, etc.)
+- [ ] Check Prax logs in system_logs/ for detailed debugging info
+- [ ] Manual feature tests at 90% stage
+
+---
+
+## When to Test What
+
+**During development (AI internal verification):**
+- Module starts without errors
+- Basic commands execute
+- Expected output appears
+
+**At 90% stage (human testing):**
+- Feature functionality (does it do what it's supposed to?)
+- Edge cases (what breaks it?)
+- Error handling (are errors honest?)
+- Integration (does it work with other modules?)
+
+**Before considering "done":**
+- Clean passes on major features
+- Errors tell the truth (no silent failures)
+- JSON logs show operations correctly
+- Acceptable standard reached (basic or advanced)
+
+---
+
+## Summary
+
+**Current approach:** Manual testing with JSON/log infrastructure
+
+**Why it works:**
+- Fast iteration without test maintenance
+- Linux environment = reliable verification
+- Error-first debugging = effective bug fixing
+- JSON/log system = verification layer
+
+**Build process:** Plan → Build to 90% → Review/test → Debug (errors first, then bugs) → Iterate → Acceptable
+
+**Future:** pytest framework when system stabilizes
+
+**Philosophy:** Build fast, verify as you go, handle errors honestly, iterate rapidly. Test infrastructure comes after stability.
+
+---
+
+## Pytest Test Structure
+
+**Current implementation:**
+```
+pytest.ini                          # Test configuration (repo root)
+tests/                              # Root test directory
+  conftest.py                       # Shared fixtures
+src/aipass/
+  api/tests/                        # API tests (4 files operational)
+    test_api_system.py
+    test_openrouter_key.py
+    test_free_models_quick.py
+    test_paid_model.py
+  prax/tests/                       # Prax tests
+    test_log_rotation.py
+  cli/tests/                        # CLI tests (infrastructure ready)
+  seedgo/
+    tests/conftest.py
+    apps/modules/test_cli_errors.py  # Demo module
+```
+
+**Running tests:**
+```bash
+# Run all tests
+pytest
+
+# Run specific branch tests
+pytest src/aipass/api/tests/
+
+# Run with markers
+pytest -m unit
+pytest -m integration
+pytest -m slow
+```
+
+**Test demonstration module:**
+- `src/aipass/seedgo/apps/modules/test_cli_errors.py` - Shows error handling patterns
+- Not a pytest test, but demonstrates testing concepts
+- Run directly: `python3 src/aipass/seedgo/apps/modules/test_cli_errors.py`
+
+---
+
+## Comments
+
+#@comments:2025-11-13:claude: Pytest infrastructure exists and operational in API/Prax branches. Selective testing approach: automate stable components, manual testing for rapid development.
+
+#@comments:2025-11-13:claude: Prax doesn't have real-time "watcher" command for logs - uses file-based logging in system_logs/ (Prax manages the location). Updated documentation to reflect actual capabilities.
+
+#@comments:2025-11-13:claude: Error-first debugging approach is critical - maybe this should be emphasized in error_handling.md when we fill that section?
+
+#@comments:2025-11-13:claude: "90% threshold" is interesting pattern - not 100% perfection, but "acceptable standard" (basic or advanced). Reflects pragmatic development philosophy.