@massu/core 0.6.0 → 0.6.2

package/commands/massu-golden-path/references/approval-points.md

@@ -44,7 +44,7 @@ Total Items: [N]
 Phases: [list]
 
 Requirements Coverage: [X]/10 dimensions resolved
-Feasibility: VERIFIED (DB, files, patterns, security)
+Feasibility: VERIFIED (config, files, patterns, security)
 Audit Passes: {iteration} (final pass: 0 gaps)
 --------------------------------------------------------------------------
 
@@ -75,7 +75,7 @@ Existing patterns checked:
 PROPOSED NEW PATTERN:
 --------------------------------------------------------------------------
 Name: [Pattern Name]
-Domain: [UI/Database/Auth/etc.]
+Domain: [Config/MCP/Hook/etc.]
 
 WRONG: [code]
 CORRECT: [code]
@@ -108,15 +108,12 @@ VERIFICATION RESULTS:
   Pattern scanner: Exit 0
   Type check: 0 errors
   Build: Exit 0
-  Lint: Exit 0
-  Prisma: Valid
+  Tests: ALL pass
+  Hook compilation: Exit 0
+  Generalization: Exit 0
   Security: No secrets staged, no credentials in code
-  VR-RENDER: All UI components rendered
-  VR-COUPLING: All backend features exposed in UI
-  VR-COLOR: No hardcoded Tailwind colors
+  Tool registration: All new tools wired
   Plan Coverage: [X]/[X] = 100%
-  Database: All environments verified
-  Help site: UP TO DATE / N/A
   Quality Score: [X.X]/5.0
 --------------------------------------------------------------------------
 
@@ -161,8 +158,8 @@ Files changed: [N] | +[N] / -[N]
 Branch: [branch] -> origin
 
 Tier 1 (Quick): PASS
-Tier 2 (Tests): PASS -- Unit: X/X, E2E: X/X, Regression: 0
-Tier 3 (Security): PASS -- Audit: 0 high/crit, RLS: verified, Secrets: clean
+Tier 2 (Tests): PASS -- Unit: X/X, Regression: 0
+Tier 3 (Security): PASS -- Audit: 0 high/crit, Secrets: clean
 --------------------------------------------------------------------------
 
 OPTIONS:
@@ -200,7 +197,7 @@ COMPETITIVE SCORECARD:
 NOTABLE DIFFERENCES:
   [Aspect]: Agent A did [X], Agent B did [Y]
 
-RECOMMENDATION: Agent {X} ({bias}) — [reason]
+RECOMMENDATION: Agent {X} ({bias}) -- [reason]
 --------------------------------------------------------------------------
 
 PER-AGENT NOTES:

package/commands/massu-golden-path/references/competitive-mode.md

@@ -1,8 +1,10 @@
 # Competitive Mode Protocol
 
+> **Shared rules apply.** Read .claude/commands/_shared-preamble.md before proceeding.
+
 > Reference doc for `/massu-golden-path --competitive`. Return to main file for overview.
 
-**Purpose**: Spawn 2-3 competing implementations of the same plan with different optimization biases, score all implementations, and select the winner before proceeding with Massu's verification rigor.
+**Purpose**: Spawn 2-3 competing implementations of the same plan with different optimization biases, score all implementations, and select the winner before proceeding with verification rigor.
 
 **Triggering**: Only when `/massu-golden-path --competitive` is explicitly used. Never automatic.
 
@@ -17,7 +19,7 @@ SCAN plan for:
   - Items with type = MIGRATION
   - Items containing ALTER TABLE, CREATE TABLE, DROP TABLE
   - Items containing RLS policies or grants
-  - Items referencing database migrations
+  - Items referencing all database environments
 
 IF any found:
   ABORT competitive mode with message:
@@ -25,7 +27,7 @@ IF any found:
    Apply migrations first, then re-run with --competitive."
 ```
 
-This mirrors the `/massu-batch` DB guard pattern (`scripts/batch-db-guard.sh`).
+This mirrors the `/massu-batch` DB guard pattern.
 
 ---
 
@@ -75,8 +77,8 @@ IMPLEMENTATION RULES:
 1. Read the plan from disk and implement ALL items
 2. Follow ALL CLAUDE.md patterns (ctx.db, protectedProcedure, etc.)
 3. Do NOT run database migrations (handled separately)
-4. Run pattern-scanner after each file: ./scripts/pattern-scanner.sh
-5. Run tsc after implementation: npx tsc --noEmit
+4. Run pattern-scanner after each file: bash scripts/massu-pattern-scanner.sh
+5. Run tsc after implementation: cd packages/core && npx tsc --noEmit
 6. Fix any issues before declaring done
 
 OUTPUT FORMAT (at completion):
@@ -249,8 +251,8 @@ IF cleanup fails:
 ### Post-Merge Verification
 
 ```
-1. Run ./scripts/pattern-scanner.sh (exit 0 required)
-2. Run npx tsc --noEmit (0 errors required)
+1. Run bash scripts/massu-pattern-scanner.sh (exit 0 required)
+2. Run cd packages/core && npx tsc --noEmit (0 errors required)
 3. IF either fails:
      Fix issues from merge
      Re-run verification

package/commands/massu-golden-path/references/error-handling.md

@@ -1,5 +1,7 @@
 # Error Handling
 
+> **Shared rules apply.** Read .claude/commands/_shared-preamble.md before proceeding.
+
 > Reference doc for `/massu-golden-path`. Return to main file for overview.
 
 ## Recoverable Errors
@@ -54,11 +56,11 @@ TO RESUME:
 
 ---
 
-## Post-Compaction Re-Verification (CR-42)
+## Post-Compaction Re-Verification (CR-12)
 
 **After ANY context compaction during a golden path run**, BEFORE continuing implementation:
 
-1. **Re-read the FULL plan document** from disk (CR-5 -- never from memory)
+1. **Re-read the FULL plan document** from disk (never from memory)
 2. **Diff every completed item against actual code**: For each item marked complete in the tracking table, re-run its VR-* verification command
 3. **VR-SPEC-MATCH audit**: For every completed UI item with specific CSS classes/structure in the plan, grep for those EXACT strings in the implementation
 4. **Flag mismatches**: Any item where implementation doesn't match the plan's exact spec -> mark as gap, fix before continuing

package/commands/massu-golden-path/references/phase-0-requirements.md

@@ -8,9 +8,9 @@
 [GOLDEN PATH -- PHASE 0: REQUIREMENTS & CONTEXT]
 ```
 
-- Call `massu_memory_sessions` for recent session context
-- Call `massu_memory_search` + `massu_memory_failures` with feature keywords
 - Read `session-state/CURRENT.md` for any prior state
+- Read `massu.config.yaml` for project configuration
+- Search memory files for relevant prior context
 
 ## 0.2 Requirements Coverage Map
 
@@ -20,7 +20,7 @@ Initialize ALL dimensions as `pending`:
 |---|-----------|--------|-------------|
 | D1 | Problem & Scope | pending | User request + interview |
 | D2 | Users & Personas | pending | Interview |
-| D3 | Data Model | pending | Phase 1A (DB Reality Check) |
+| D3 | Data Model | pending | Phase 1A (Config/Schema Reality Check) |
 | D4 | Backend / API | pending | Phase 1A (Codebase Reality Check) |
 | D5 | Frontend / UX | pending | Interview + Phase 1A |
 | D6 | Auth & Permissions | pending | Phase 1A (Security Pre-Screen) |

package/commands/massu-golden-path/references/phase-1-plan-creation.md

@@ -12,85 +12,75 @@
 
 ### 1A.1 Feature Understanding
 
-- Call `massu_knowledge_search`, `massu_knowledge_pattern`, `massu_knowledge_schema_check` with feature name
 - Document: exact user request, feature type, affected domains
-- Search codebase for similar features, routers, pages
+- Search codebase for similar features, tool modules, existing patterns
+- Read `massu.config.yaml` for relevant config sections
 
-### 1A.2 Database Reality Check (VR-SCHEMA-PRE)
+### 1A.2 Config & Schema Reality Check
 
-For EACH table the feature might use, query via MCP:
+For features touching config or databases:
 
-```sql
-SELECT column_name, data_type, is_nullable, column_default
-FROM information_schema.columns WHERE table_name = '[TABLE]' ORDER BY ordinal_position;
+- Parse `massu.config.yaml` and verify all referenced config keys exist
+- Check SQLite schema for affected tables (`getCodeGraphDb`, `getDataDb`, `getMemoryDb`)
+- Verify tool definitions in `tools.ts` for any tools being modified
 
-SELECT polname, polcmd FROM pg_policies WHERE tablename = '[TABLE]';
+Document: existing config keys, required new keys, required schema changes.
 
-SELECT grantee, privilege_type FROM information_schema.table_privileges WHERE table_name = '[TABLE]';
-```
-
-Run `./scripts/check-bad-columns.sh`. Call `massu_schema` for Prisma cross-reference.
-Document: existing tables, required new tables/columns, migration SQL previews.
+### 1A.3 Config-Code Alignment (VR-CONFIG)
 
-### 1A.3 Config-Code Alignment (VR-DATA)
+If feature uses config-driven values:
 
-If feature uses DB-stored configs:
-
-```sql
-SELECT DISTINCT jsonb_object_keys(config_column) as keys FROM config_table;
+```bash
+# Check config keys used in code
+grep -rn "getConfig()" packages/core/src/ | grep -o 'config\.\w\+' | sort -u
+# Compare to massu.config.yaml structure
 ```
 
-Compare to code: `grep -rn "config\." src/lib/[feature]/ | grep -oP 'config\.\w+' | sort -u`
-
 ### 1A.4 Codebase Reality Check
 
 - Verify target directories/files exist
-- Read similar routers and components
-- Load relevant pattern files (database/auth/ui/realtime/build)
+- Read similar tool modules and handlers
+- Load relevant pattern files (build/testing/security/database/mcp)
 
 ### 1A.5 Blast Radius Analysis (CR-25)
 
-**MANDATORY when plan changes any constant, path, route, enum, or config key.**
+**MANDATORY when plan changes any constant, export name, config key, or tool name.**
 
 1. Identify ALL changed values (old -> new)
 2. Codebase-wide grep for EACH value
-3. Call `massu_impact` for indirect impact through import chains
-4. If plan deletes files: call `massu_sentinel_impact` -- zero orphaned features allowed
-5. Categorize EVERY occurrence: CHANGE / KEEP (with reason) / INVESTIGATE
-6. Resolve ALL INVESTIGATE to 0. Add ALL CHANGE items as plan deliverables.
+3. If plan deletes files: verify no remaining imports or references
+4. Categorize EVERY occurrence: CHANGE / KEEP (with reason) / INVESTIGATE
+5. Resolve ALL INVESTIGATE to 0. Add ALL CHANGE items as plan deliverables.
 
 ### 1A.6 Pattern Compliance Check
 
-Check applicable patterns: ctx.db, user_profiles, 3-step query, BigInt/Decimal, RLS+Grants, Suspense, Select.Item, protectedProcedure, Zod validation. Read most similar router/component for patterns used.
+Check applicable patterns: ESM imports (.ts extensions), config access (getConfig()), tool registration (3-function pattern), hook compilation (esbuild), SQLite DB access (getCodeGraphDb/getDataDb/getMemoryDb), memDb lifecycle (try/finally close).
 
-### 1A.7 Backend-Frontend Coupling Check (CR-12)
+Read most similar tool module for patterns used.
 
-For EVERY backend z.enum, type, or procedure planned -- verify a corresponding frontend item exists. If NOT, ADD IT.
+### 1A.7 Tool Registration Check
+
+For EVERY new MCP tool planned -- verify a corresponding registration item exists in the plan (definitions + routing + handler in `tools.ts`). If NOT, ADD IT.
 
 ### 1A.8 Question Filtering
 
 1. List all open questions
-2. Self-answer anything answerable by reading code or querying DB
+2. Self-answer anything answerable by reading code or config
 3. Surface only business logic / UX / scope / priority questions to user via AskUserQuestion
 4. If all self-answerable, skip user prompt
 
-### 1A.9 Security Pre-Screen (6 Dimensions)
+### 1A.9 Security Pre-Screen (5 Dimensions)
 
 | Dim | Check | If Triggered |
 |-----|-------|-------------|
-| S1 | PII / Sensitive Data | Add RLS + column-level access |
-| S2 | Authentication | Verify protectedProcedure |
-| S3 | Authorization | Add RBAC checks, RLS policies |
-| S4 | Injection Surfaces | Add Zod validation, parameterized queries |
-| S5 | Secrets Management (CR-5) | Add AWS Secrets Manager items |
-| S6 | Rate Limiting | Add rate limiting middleware |
+| S1 | PII / Sensitive Data | Add access controls |
+| S2 | Authentication | Verify auth checks |
+| S3 | Authorization | Add permission checks |
+| S4 | Injection Surfaces | Add input validation, parameterized queries |
+| S5 | Rate Limiting | Add rate limiting considerations |
 
 **BLOCKS_REMAINING must = 0 before proceeding.**
 
-### 1A.10 ADR Generation (Optional)
-
-For architectural decisions: `massu_adr_list` -> `massu_adr_generate`.
-
 Mark all coverage dimensions as `done` or `n/a`.
 
 ---
@@ -101,25 +91,24 @@ Mark all coverage dimensions as `done` or `n/a`.
 [GOLDEN PATH -- PHASE 1B: PLAN GENERATION]
 ```
 
-Write plan to: `plans/[YYYY-MM-DD]-[feature-name].md`
+Write plan to: `docs/plans/[YYYY-MM-DD]-[feature-name].md`
 
 **Plan structure** (P-XXX numbered items):
 - Overview (feature, complexity, domains, item count)
 - Requirements Coverage Map (D1-D10 all resolved)
-- Phase 0: Credentials & Secrets (CR-5)
-- Phase 1: Database Changes (migrations with exact SQL)
-- Phase 2: Backend Implementation (routers, procedures, input schemas)
-- Phase 3: Frontend Implementation (components, pages, renders-in)
+- Phase 1: Configuration Changes (massu.config.yaml)
+- Phase 2: Backend Implementation (tool modules, handlers, SQLite schema)
+- Phase 3: Frontend/Hook Implementation (hooks, plugin code)
 - Phase 4: Testing & Verification
-- Phase 5: Documentation (help site pages, changelog)
+- Phase 5: Documentation
 - Verification Commands table
 - Item Summary table
 - Risk Assessment
 - Dependencies
 
-**Item numbering**: P0-XXX (secrets), P1-XXX (database), P2-XXX (backend), P3-XXX (frontend), P4-XXX (testing), P5-XXX (docs).
+**Item numbering**: P1-XXX (config), P2-XXX (backend), P3-XXX (frontend/hooks), P4-XXX (testing), P5-XXX (docs).
 
-**Implementation Specificity Check**: Every item MUST have exact file path, exact content/SQL, insertion point, format matches target, verification command.
+**Implementation Specificity Check**: Every item MUST have exact file path, exact content, insertion point, format matches target, verification command.
 
 **Documentation Impact Assessment**: If ANY user-facing features, Phase 5 deliverables are MANDATORY.
 
@@ -141,7 +130,7 @@ WHILE true:
   result = Task(subagent_type="massu-plan-auditor", model="opus", prompt="
     Audit iteration {iteration} for plan: {PLAN_PATH}
     Execute ONE complete audit pass. Verify ALL deliverables.
-    Check: VR-PLAN-FEASIBILITY, VR-PLAN-SPECIFICITY, Pattern Alignment, Schema Reality.
+    Check: VR-PLAN-FEASIBILITY, VR-PLAN-SPECIFICITY, Pattern Alignment, Config Reality.
     Fix any plan document gaps you find.
 
     CRITICAL: Report GAPS_DISCOVERED as total gaps FOUND, EVEN IF you fixed them.
@@ -157,7 +146,7 @@ WHILE true:
 END WHILE
 ```
 
-**VR-PLAN-FEASIBILITY**: DB schema exists, files exist, dependencies available, patterns documented, credentials planned.
+**VR-PLAN-FEASIBILITY**: Files exist, config keys valid, dependencies available, patterns documented.
 **VR-PLAN-SPECIFICITY**: Every item has exact path, exact content, insertion point, verification command.
 **Pattern Alignment**: Cross-reference ALL applicable patterns from CLAUDE.md and patterns/*.md.
 

package/commands/massu-golden-path/references/phase-2-implementation.md

@@ -21,12 +21,12 @@ ELSE:
 [GOLDEN PATH -- PHASE 2: IMPLEMENTATION]
 ```
 
-1. Read plan from disk (NOT memory)
+1. Read plan from disk (NOT memory -- CR-5)
 2. Extract ALL deliverables into tracking table:
 
 | Item # | Type | Description | Location | Verification | Status |
 |--------|------|-------------|----------|--------------|--------|
-| P1-001 | MIGRATION | ... | ... | VR-SCHEMA | PENDING |
+| P1-001 | CONFIG | ... | ... | VR-CONFIG | PENDING |
 
 3. Create VR-PLAN verification strategy:
 
@@ -38,38 +38,12 @@ ELSE:
 
 ---
 
-## Phase 2A.5: Sprint Contracts
-
-> Full protocol: [sprint-contract-protocol.md](sprint-contract-protocol.md)
-
-**Before implementation begins**, negotiate a sprint contract for each plan item:
-
-1. For each plan item in the tracking table:
-   - Define **Scope Boundary** (IN/OUT)
-   - Define **Implementation Approach** (files, patterns)
-   - Write **3-5 Acceptance Criteria** (must be specific enough that two independent evaluators agree on PASS/FAIL)
-   - Map to **VR-\* Verification Types**
-
-2. Add contract columns to the Phase 2A tracking table:
-
-| Item # | Type | Description | Location | Verification | Scope Boundary | Acceptance Criteria | Contract Status |
-|--------|------|-------------|----------|--------------|----------------|---------------------|-----------------|
-| P1-001 | MIGRATION | ... | ... | VR-SCHEMA | IN: ... / OUT: ... | 1. ... 2. ... 3. ... | AGREED |
-
-3. **Quality bar**: Criteria using words like "good", "correct", "proper" without specifics = reject and rewrite. Each contract must include criteria from at least 3 categories: happy path, data display, empty/loading/error states, user feedback, edge cases.
-
-4. **Skip conditions**: Mark `Contract: N/A` for pure refactors (VR-BUILD + VR-TYPE + VR-TEST sufficient), documentation-only items, and migrations where SQL IS the contract.
-
-5. **Max 3 negotiation rounds** per item. If unresolved, escalate via AskUserQuestion.
-
----
-
 ## Phase 2B: Implementation Loop
 
 For each plan item:
-1. **Pre-check**: Load CR rules, domain patterns for the affected file
+1. **Pre-check**: Verify file exists, read current state
 2. **Execute**: Implement the item following established patterns
-3. **Guardrail**: Run `./scripts/pattern-scanner.sh` (ABORT if fails)
+3. **Guardrail**: Run `bash scripts/massu-pattern-scanner.sh` (ABORT if fails)
 4. **Verify**: Run applicable VR-* checks with proof
 5. **VR-PIPELINE**: If the item involves a data pipeline (AI, cron, generation, ETL), trigger the pipeline manually, verify output is non-empty. Empty output = fix before continuing.
 6. **Update**: Mark item complete in tracking table
@@ -83,16 +57,13 @@ For each plan item:
 
 ```
 CHECKPOINT:
-[1] READ plan section    [2] QUERY DB             [3] GREP routers
-[4] LS components        [5] VR-RENDER check      [6] VR-COUPLING check
-[7] Pattern scanner      [8] npm run build         [9] npx tsc --noEmit
-[10] npm run lint        [11] npx prisma validate  [12] npm test
-[13] UI/UX verification  [14] API/router verification  [15] Security check
-[16] COUNT gaps -> IF > 0: FIX and return to [1]
+[1] READ plan section          [2] GREP tool registrations    [3] LS modules
+[4] VR-CONFIG check            [5] VR-TOOL-REG check          [6] VR-HOOK-BUILD check
+[7] Pattern scanner            [8] npm run build               [9] cd packages/core && npx tsc --noEmit
+[10] npm test                  [11] VR-GENERIC check           [12] Security scanner
+[13] COUNT gaps -> IF > 0: FIX and return to [1]
 ```
 
-> **Cross-reference**: Full checkpoint audit protocol with detailed steps is in `massu-loop/references/checkpoint-audit.md`.
-
 ---
 
 ## Phase 2C: Multi-Perspective Review
@@ -112,60 +83,14 @@ architecture_result = Task(subagent_type="massu-architecture-reviewer", model="o
   Return structured result with ARCHITECTURE_GATE: PASS/FAIL.
 ")
 
-ux_result = Task(subagent_type="massu-ux-reviewer", model="sonnet", prompt="
+quality_result = Task(subagent_type="massu-quality-reviewer", model="sonnet", prompt="
   Review implementation for plan: {PLAN_PATH}
-  Focus: UX, accessibility, loading/error/empty states, consistency.
-  Return structured result with UX_GATE: PASS/FAIL.
+  Focus: Code quality, ESM compliance, config-driven patterns, TypeScript strict mode, test coverage.
+  Return structured result with QUALITY_GATE: PASS/FAIL.
 ")
 ```
 
-**Phase 2C.2: QA Evaluator** (conditional -- UI plans only)
-
-> Full spec: [qa-evaluator-spec.md](qa-evaluator-spec.md)
-
-If the plan touches UI files, spawn an adversarial QA evaluator:
-
-```
-IF plan has UI files:
-  qa_result = Task(subagent_type="massu-ux-reviewer", model="opus", prompt="
-    === QA EVALUATOR MODE ===
-    You are an ADVERSARIAL QA agent. Your job is to FIND BUGS, not approve work.
-
-    Plan: {PLAN_PATH}
-    Sprint contracts: {CONTRACTS_FROM_2A5}
-
-    For EACH plan item with a sprint contract:
-    1. NAVIGATE to the affected page using Playwright MCP
-    2. EXERCISE the feature as a real user would
-    3. VERIFY against sprint contract acceptance criteria (EVERY criterion)
-    4. CHECK for known failure patterns:
-       - Mock/hardcoded data (data doesn't change when DB changes)
-       - Write succeeds but read/display broken
-       - Feature stubs (onClick/onSubmit empty or log-only)
-       - Invisible elements (display:none, opacity:0, z-index buried)
-       - Missing query invalidation (create item, verify list updates without refresh)
-    5. GRADE: PASS / PARTIAL / FAIL with specific evidence
-
-    ANTI-LENIENCY RULES:
-    - Never say 'this is acceptable because...' — if criteria aren't met, it's FAIL
-    - Never give benefit of the doubt — if you can't verify it works, it's FAIL
-    - Partial credit is still failure — PARTIAL means 'not done yet'
-    - Every PASS must cite specific evidence (screenshot, DOM state, network response)
-
-    Return structured result with QA_GATE: PASS/FAIL and per-item grades.
-  ")
-ELSE:
-  Log: "QA Evaluator: SKIPPED (no UI files in plan)"
-```
-
-**Gate logic**: Fix ALL CRITICAL/HIGH findings before proceeding. WARN findings = document and proceed.
-
-```
-GATES = [SECURITY_GATE, ARCHITECTURE_GATE, UX_GATE]
-IF plan has UI files: GATES += [QA_GATE]
-IF ANY gate == FAIL: Fix findings and re-run failed gates
-ALL gates must PASS before proceeding to Phase 2D.
-```
+Fix ALL findings at ALL severity levels before proceeding (CR-45). CRITICAL, HIGH, MEDIUM, LOW — all get fixed. No severity is exempt.
 
 ---
 
@@ -176,36 +101,15 @@ iteration = 0
 WHILE true:
   iteration += 1
 
-  # Circuit breaker (detect stagnation)
-  IF iteration >= 3:
-    stalled_items = items that failed in ALL of last 3 iterations
-    IF stalled_items.length > 0:
-      Log: "REFINE-OR-PIVOT: {stalled_items.length} items stalled for 3+ iterations"
-      FOR EACH stalled_item:
-        IF same_root_cause_each_time: REFINE (targeted fix for root cause)
-        IF different_failures_each_time: PIVOT (scrap approach, try alternative)
-        IF no_clear_pattern: AskUserQuestion with evidence from last 3 attempts
+  # Circuit breaker (CR-37)
+  IF iteration >= 3 AND same gaps as previous iteration:
+    AskUserQuestion: "Loop stalled after {iteration} passes. Re-plan / Continue / Stop?"
 
   result = Task(subagent_type="massu-plan-auditor", model="opus", prompt="
     Audit iteration {iteration} for plan: {PLAN_PATH}
     Verify ALL deliverables with VR-* proof.
     Check code quality (patterns, build, types, tests).
     Check plan coverage (every item verified).
-
-    VR-SPEC-MATCH: For EVERY UI plan item with specific CSS classes,
-    component names, or layout instructions -- grep the implementation for those
-    EXACT strings. Missing = gap.
-
-    VR-PIPELINE: For features with data pipelines (AI, cron, generation),
-    trigger the pipeline procedure and verify output is non-empty. Empty = gap.
-
-    SPRINT CONTRACT VERIFICATION: For each plan item with a sprint contract
-    (from Phase 2A.5), verify EVERY acceptance criterion is met:
-    - Read the contract's acceptance criteria list
-    - Test each criterion with specific evidence (screenshot, grep, DOM state)
-    - Any unmet criterion = gap, even if the code 'looks right'
-    - Contract criteria are IN ADDITION TO VR-* checks — both must pass
-
     Fix any gaps you find.
 
     CRITICAL: GAPS_DISCOVERED = total FOUND, even if fixed.
@@ -222,7 +126,7 @@ END WHILE
 
 ---
 
-## Phase 2E: Post-Build Reflection + Memory Persist
+## Phase 2E: Post-Build Reflection + Memory Persist (CR-38)
 
 **MANDATORY -- reflection + memory write = ONE atomic action.**
 
@@ -240,14 +144,13 @@ Apply any low-risk refactors immediately. Log remaining suggestions in plan unde
 
 ## Phase 2F: Documentation Sync (User-Facing Features)
 
-If plan includes ANY user-facing features:
+If plan includes ANY user-facing features (new MCP tools, config changes, hook changes):
 
-1. Audit documentation against code changes
-2. Update affected documentation pages
-3. Add changelog entry
-4. Commit documentation updates (separate repo if applicable)
+1. Update relevant documentation (README, API docs, config docs)
+2. Ensure tool descriptions match implementation
+3. Update config schema documentation if config keys changed
 
-Skip ONLY if purely backend/infra with zero user-facing changes.
+Skip ONLY if purely internal refactoring with zero user-facing changes.
 
 ---
 
@@ -257,28 +160,27 @@ Skip ONLY if purely backend/infra with zero user-facing changes.
 [GOLDEN PATH -- PHASE 2G: BROWSER VERIFICATION]
 ```
 
-**Auto-trigger condition**: If plan touches ANY UI files, this phase runs automatically. If purely backend/infra with zero UI changes, skip with log note: `Browser verification: SKIPPED (no UI files changed)`.
-
-**SAFETY RULE**: NEVER use real client data. NEVER click destructive actions (Delete, Send, Submit) on production.
+**Auto-trigger condition**: If plan touches ANY UI/demo files or produces visual output, this phase runs automatically. If purely backend/MCP/config with zero visual output, skip with log note: `Browser verification: SKIPPED (no UI files changed)`.
 
 ### 2G.1 Determine Target Pages
 
-Map changed files to URLs:
+Map changed features to testable URLs:
+- If the project has a demo page or documentation site: test affected pages
+- If testing MCP tool output: use a test harness or verify tool responses
 - Component changes: identify ALL pages that render the component
-- Layout changes: test ALL child routes under that layout
 
 ### 2G.2 Browser Setup & Authentication
 
-Use Playwright MCP plugin tools.
+Use Playwright MCP plugin tools (`mcp__plugin_playwright_playwright__*`). Fallback: `mcp__playwright__*`.
 
-1. `browser_navigate` to first target URL
-2. `browser_snapshot` to check auth status
-3. If redirected to `/login` or auth check visible: STOP and request manual login
+1. `browser_navigate` to target URL
+2. `browser_snapshot` to check page status
+3. If authentication required: STOP and request manual login
 
 ```
 AUTHENTICATION REQUIRED
 
-The Playwright browser is not logged in to the app.
+The Playwright browser is not logged in to the target application.
 Please log in manually in the open browser window, then re-run the golden path.
 ```
 
@@ -290,24 +192,24 @@ For EACH target page:
 
 | Check | Tool | Captures |
 |-------|------|----------|
-| Console errors/warnings | `browser_console_messages` | React errors, TypeError, CSP violations, auth warnings |
+| Console errors/warnings | `browser_console_messages` | React errors, TypeError, CSP violations |
 | Network failures | `browser_network_requests` | 500s, 404s, CORS failures, timeouts |
 
 Categorize findings:
 
 | Category | Severity |
 |----------|----------|
-| React crash, 500 error, data exposure | **P0 -- CRITICAL** |
-| Network failure, CSP violation, broken interaction, auth warning | **P1 -- HIGH** |
-| Visual issues, performance warnings, broken images | **P2 -- MEDIUM** |
-| Console warnings, deprecations, i18n missing keys | **P3 -- LOW** |
+| Crash, 500 error, data exposure | **P0 -- CRITICAL** |
+| Network failure, broken interaction | **P1 -- HIGH** |
+| Visual issues, performance warnings | **P2 -- MEDIUM** |
+| Console warnings, deprecations | **P3 -- LOW** |
 
 ### 2G.4 Interactive Testing (Per Page)
 
-1. `browser_snapshot` -> inventory ALL interactive elements
+1. `browser_snapshot` -> inventory ALL interactive elements (buttons, links, forms, selects, tabs, modals, data tables)
 2. For EACH testable element:
-   - Capture console state BEFORE interaction
-   - Perform interaction
+   - Capture console state BEFORE interaction (`browser_console_messages`)
+   - Perform interaction (`browser_click`, `browser_select_option`, `browser_fill_form`)
    - Wait 2-3 seconds for async operations
    - Capture console state AFTER interaction
    - Record any NEW errors introduced
@@ -319,15 +221,15 @@ Categorize findings:
 ### 2G.5 Visual & Performance Audit
 
 **Visual checks**:
-- Broken images: find `img` elements with `naturalWidth === 0`
+- Broken images: `browser_evaluate` to find `img` elements with `naturalWidth === 0`
 - Layout issues: overflow, overlapping, missing content, broken alignment
-- Responsive: test at 1440x900 (desktop), 768x1024 (tablet), 375x812 (mobile)
-- Screenshot evidence at each breakpoint if issues found
+- Responsive: `browser_resize` at 1440x900 (desktop), 768x1024 (tablet), 375x812 (mobile)
+- Screenshot evidence: `browser_take_screenshot` at each breakpoint if issues found
 
 **Performance checks**:
-- Page load timing
-- Resources > 500KB
-- Slow API calls > 3s, duplicate requests
+- Page load timing via `browser_evaluate` (`performance.getEntriesByType('navigation')`)
+- Resources > 500KB via `browser_evaluate` (`performance.getEntriesByType('resource')`)
+- Slow API calls > 3s, duplicate requests via `browser_network_requests`
 
 | Metric | Good | Needs Work | Critical |
 |--------|------|------------|----------|
@@ -361,10 +263,9 @@ Report includes: summary table, console errors, network failures, interactive el
 ### 2G.8 Auto-Learning Protocol
 
 For EACH browser-discovered fix:
-1. Record with type="bugfix" including browser symptom -> code fix mapping
-2. Update MEMORY.md with symptom/root cause/fix/files
-3. Add to `scripts/pattern-scanner.sh` if the bad pattern is grep-able
-4. Codebase-wide search for same bad pattern (CR-9) -- fix ALL instances
+1. Update memory files with symptom/root cause/fix/files
+2. Add to `scripts/massu-pattern-scanner.sh` if the bad pattern is grep-able
+3. Codebase-wide search for same bad pattern (CR-9) -- fix ALL instances
 
 ---
 
@@ -386,11 +287,9 @@ The golden path spawns multiple subagents across Phase 2. Follow these principle
 
 ```
 [GOLDEN PATH -- PHASE 2 COMPLETE]
-  Sprint contracts: NEGOTIATED ({N} items contracted, {M} N/A)
   All plan items implemented
-  Multi-perspective review: PASSED (security, architecture, UX)
-  QA evaluator: PASSED / SKIPPED (no UI files)
-  Verification audit: PASSED (Loop #{iteration}, 0 gaps, contracts verified)
+  Multi-perspective review: PASSED (security, architecture, quality)
+  Verification audit: PASSED (Loop #{iteration}, 0 gaps)
   Post-build reflection: PERSISTED to memory
   Documentation sync: COMPLETE / N/A
   Browser verification: PASSED ({N} pages tested, {M} issues fixed) / SKIPPED (no UI files)