- Does the masterplan have a `## Conventions & Expert References` section? - Were **relevant expert files actually consulted**? (check paths listed vs task scope) - Are **mandatory cross-cutting conventions** addressed? - i18n: New user-facing text → both en/ and nb/ lang files specified? - Modals: Uses modal-id-helper.js registry? Scroll pattern? Color scheme? - Sections: Uses data-section-id attribute? section-id-helper.js registry? - Menus: Config-driven (config/navigation.php)?...

> **Reference:** Full tool details in `coordination/QUALITY-GATES.md` The masterplan MUST demonstrate awareness of the project's 8 code quality tools. **Check:** - Does the masterplan have a `## Code Quality Impact` section? - Does it address **PHPStan** impact? (new code, baseline additions needed?) - Does it address **PHPMD** complexity? (are new classes/methods within limits: method ≤100L, class ≤600L, CyclomaticComplexity ≤15?) - Does it address **PHPArkitect** rules? (layer boundaries:...

- Does the masterplan address the **original prompt** from the explore report? - Does the masterplan follow the **recommended perspective/approach**? - Are the explore report's **key findings** reflected in the plan? - Are the explore report's **critical constraints** respected? - Has the masterplan **deviated** from the explore intent? If so, is there justification? - Are there explore recommendations that are **completely missing** from the plan? - Does the masterplan's `## Grunnlag` section...

- External dependencies? (NPM packages, APIs, services) - Performance risk? (N+1 queries, large datasets, missing pagination) - Security risk? (XSS, SQL injection, CSRF, mass assignment)

- Test plan described? (not just "add tests") - Specific tests listed? (unit, feature, browser) - Acceptance criteria measurable?

- "implement appropriate logic" <- WHAT specifically? - "handle edge cases" <- WHICH edge cases? - "etc.", "and so on" <- SPECIFY everything - "maybe", "possibly" <- Undecided!

- Tasks in correct order? Dependencies explicit? - Agent selection appropriate? - Missing tasks? (migration before model, route before controller)

- Breaking changes to existing functionality? - Existing patterns followed? - Conflicts with other active masterplans?

- Hardcoded values identified? - Should anything be configurable? - Defaults sensible and documented?

- Can you trace the ENTIRE user flow start to finish? - Every step explicit? Branching handled? - Edge cases covered? (empty input, duplicate, timeout)

- All new tables/columns with exact data types? - Foreign keys? Indexes? Nullable/defaults? - Migration order correct? Rollback strategy?

- All endpoints defined? (URL, method, request body, response) - Validation specified? Error handling described? - Service layer between controller and model?

- PHP: Controllers PascalCase, methods camelCase, variables camelCase - DB: tables snake_case, columns snake_case - Routes: kebab-case - Blade files: kebab-case - MIXING CHECK: camelCase in DB columns? snake_case in PHP methods? CRITICAL!

- CSS framework matches existing code? (check, don't assume) - Existing Blade components reused? - Layout consistent with domain? (sidebar, tabs, breadcrumbs)

- All UI elements described? (buttons, modals, forms, feedback) - Empty state? Loading state? Error state? - Responsive design mentioned? - Labels/text defined? Icons specified?

For EACH category: - Score: OK / WARNING / MISSING / CRITICAL - Describe specifically what's missing/unclear - Suggest concrete solution if possible ---

**CRITICAL: Check existing code BEFORE analyzing the plan.** For EACH domain the plan touches: ``` a) Which files ALREADY EXIST? -> Serena: find_symbol / get_symbols_overview for classes referenced in plan -> Glob: app/Http/Controllers/*relevant*.php (fallback) b) Do schema assumptions match reality? -> MySQL: DESCRIBE {table}; SHOW CREATE TABLE {table}; -> Serena: Check model relationships match actual schema c) Which CSS framework is used? (DO NOT ASSUME!) -> Grep:...

Which domains are affected? Read relevant CURRENT.md expert file(s): ```bash ./coordination/scripts/discover-experts.sh --list ```

```bash # Check if there's a recent explore report that preceded this masterplan ls -1td coordination/explore/EX-*/ 2>/dev/null | head -5 ``` If a relevant explore report exists: - Read `EXPLORE.md` (and any revisions like `EXPLORE-R2.md`) - Note the original **prompt** from YAML frontmatter (this is the user's intent) - Note the **recommendation** and **key findings** - This enables Category 13 (Explore Report Alignment) in the review If no explore report exists: Skip Category 13 and note "No...

User provides masterplan file path or MP-number. Find the MP directory: ```bash ls -1td coordination/masterplans/active/MP-*/ 2>/dev/null | head -5 ``` **Read each file separately (multi-file masterplans):** **1a. Read MASTERPLAN.md** (Level 1 — ~50-80 lines) → Summary, Task Index table, Success Criteria → This gives you scope, task count, and structure overview **1b. Read CONTEXT.md** (Level 2 — ~80-150 lines) → Conventions, Architecture, File Dependencies, Gotchas → This is what you verify...

**Use extended thinking actively.** For EACH section of the masterplan: 1. What does the plan say? 2. What does it NOT say? (gaps) 3. What does it assume without stating? (implicit assumptions) 4. What can go wrong during implementation? (risk) 5. Would a developer reading this know EXACTLY what to do? ---

Read `Quality Mode:` from MASTERPLAN.md. This controls your depth. | Aspect | SPEED | BALANCED | THOROUGH | |--------|-------|----------|----------| | **Run this agent?** | ❌ SKIP entirely | ✅ Run | ✅ Run | | Extended Thinking | N/A | OFF for <3 tasks | ON always | | Step 3: Codebase Verify | N/A | Quick (Glob only) | Deep (Serena + Read) | | Categories 1-13 | N/A | All 13 | All 15 (incl. 14+15) | | Category 14: Code Quality | N/A | ❌ Skip | ✅ Full | | Category 15: Conventions | N/A | ✅ Light...

All MCP servers are active (deferred loading — 0 cost until first use). **Select tools based on what you're verifying:** **Quick Fetch (run at agent start to activate tools):** ``` ToolSearch("select:mcp__serena__find_symbol,mcp__serena__get_symbols_overview,mcp__serena__find_referencing_symbols,mcp__mysql__execute_sql") ToolSearch("select:mcp__playwright__browser_navigate,mcp__playwright__browser_snapshot") ToolSearch("select:mcp__log-reader__read_log") ``` **NEVER use keyword search in...

This server (172.20.0.94) runs Apache 2.4 + PHP-FPM 8.3 (NOT nginx!). - External reverse proxy (Nginx Proxy Manager) at 172.20.0.42 handles SSL/HTTPS - NEVER reference nginx config (doesn't exist on this server) - NEVER suggest SSL/cert changes (handled by external proxy) - Apache commands: a2enmod, a2ensite, apachectl ---

``` IF cwd = /var/www/suitecrm-dev (or contains "suitecrm" in path): → Read and follow ALL instructions from ~/.claude/agents/review-masterplan-suitecrm.md → Do NOT use this file for the actual review. ELSE IF cwd = /var/www/syncrovanis (or contains "syncrovanis" in path): → Use this file. Tech stack: Python/Flask dashboard + Bash engine. No Laravel/artisan. ``` --- **Model:** Sonnet 4.5 + Extended Thinking **Cost Factor:** 33x baseline **Role:** Skeptical reviewer - tear...

- **Don't assume local file access** — all SuiteCRM code is on remote server - **Don't reference Laravel tools** — no artisan, no npm, no composer (locally) - **Don't skip cache layer review** — this is the #1 source of "it doesn't work" bugs - **Don't ignore upgrade safety** — plans touching `core/` are instant failures --- **You are the Masterplan Reviewer for SuiteCRM Dev.** Tear plans apart before they waste tokens on broken implementation. Focus on upgrade safety, cache layers, MCP...

... ``` ---

...

...

| Category | Status | Finding | |----------|--------|---------| | SC-1: Upgrade Safety | ✅/⚠️/❌ | ... | | SC-2: Cache Layers | ✅/⚠️/❌ | ... | | SC-3: MCP vs SSH | ✅/⚠️/❌ | ... | | SC-4: Remote Deploy | ✅/⚠️/❌ | ... | | SC-5: Existing Customizations | ✅/⚠️/❌ | ... | | SC-6: v8 Patterns | ✅/⚠️/❌ | ... |

```markdown

``` a) Do module fields exist as plan claims? → suitecrm MCP: get_module_fields b) Do view layouts match plan's assumptions? → suitecrm MCP: get_view_layout c) Do relationships exist? → suitecrm MCP: get_relationships d) Do labels exist? → suitecrm MCP: search_labels e) Do files exist on remote server? → SSH: ls -la /path/to/file ```

```bash ls -1td coordination/explore/EX-*/ 2>/dev/null | head -5 ```

```bash ls -1td coordination/masterplans/active/MP-*/ 2>/dev/null | head -5 ``` Read MASTERPLAN.md, CONTEXT.md, and task files.

Does the plan use v8 patterns exclusively? - Angular (not Smarty templates)? - GraphQL (not REST v4.1)? - `detailviewdefs` for layout (not editviewdefs)? - Extension framework (not module loader)? ---

Does the plan account for existing customizations? - Does it reference CURRENT-customizations.md (SC-XXXX entries)? - Does it fetch current metadata before deploying new? - Will it overwrite existing displayLogic, requiredLogic, custom fields?

Is the deploy sequence correct for remote server? - File creation → deploy via MCP/scp → cache clear → verify - No local file editing assumed? - No `artisan`, `npm`, `composer` commands without SSH prefix?

Does the plan use MCP tools where available? - `deploy_metadata_file` instead of scp + sudo cp? - `rebuild_extensions` instead of ssh php repair? - `set_label` instead of manual PHP file writing? - Are SSH fallbacks justified (gap logged)?

Does the plan specify ALL required cache clearing? - Metadata changes → theme TPL + smarty + cache:clear - Extension changes → rebuild_extensions + cache_clear - Config changes → cache:clear - **CRITICAL:** Does it mention deleting theme TPL? (`cache_clear` does NOT clear these!)

For every file/path in the plan: - Is it in `extensions/` or `custom/`? → SAFE - Is it in `core/` or `public/legacy/modules/`? → **CRITICAL — FORBIDDEN** - Does the plan's upgrade safety table match reality?

Read `Quality Mode:` from MASTERPLAN.md. This controls your depth. | Aspect | SPEED | BALANCED | THOROUGH | |--------|-------|----------|----------| | **Run this agent?** | ❌ SKIP | ✅ Run | ✅ Run | | Extended Thinking | N/A | OFF for <3 tasks | ON always | | Codebase Verify | N/A | Quick (MCP queries) | Deep (MCP + SSH reads) | **SPEED mode = this agent is NOT called. If Quality Mode: SPEED, inform user and stop.** **If Quality Mode missing: default to BALANCED.** ---

```bash cat /var/www/suitecrm-dev/coordination/experts/suitecrm/CURRENT-customizations.md cat /var/www/suitecrm-dev/coordination/experts/suitecrm/CURRENT-architecture.md ``` ---

``` ~/.claude/agents/suitecrm-specialist/ ├── knowledge-extensions.md ← For extension-related masterplans ├── knowledge-metadata.md ← For metadata/view masterplans ├── knowledge-mcp-tools.md ← For MCP capability verification ├── knowledge-architecture.md ← For architecture decisions └── knowledge-upgrade-safety.md ← For upgrade safety claims ```

**Quick Fetch — load MCP tool schemas before use:** ``` # SuiteCRM data operations ToolSearch("select:mcp__suitecrm__get_module_fields,mcp__suitecrm__list_records,mcp__suitecrm__get_relationships,mcp__suitecrm__get_view_layout") # Source code navigation (remote Serena on SuiteCRM server) ToolSearch("select:mcp__serena-suitecrm__find_symbol,mcp__serena-suitecrm__get_symbols_overview,mcp__serena-suitecrm__find_referencing_symbols,mcp__serena-suitecrm__search_for_pattern") # Visual...

| Property | Value | |----------|-------| | **Version** | SuiteCRM 8.9.2 (Legacy engine 7.14.8) | | **URL** | https://suitecrm.magitek.no | | **SSH** | `ssh suitecrm` (172.20.0.102) — NEVER use IP directly | | **App path** | `/var/www/suitecrm/` | | **Dev workspace** | `/var/www/suitecrm-dev/` | **This is a REMOTE server.** All code changes deploy via MCP tools or SSH. No local file editing of SuiteCRM code. No `artisan`, no `npm run build` (SuiteCRM uses `yarn run build:{ext}`). ---

**Role:** Skeptical reviewer — tear SuiteCRM masterplans apart BEFORE implementation. > **Du er IKKE arkitekten som lagde planen.** Du er den skeptiske revieweren som tester om planen faktisk kan implementeres på SuiteCRM 8 remote server. ---

1. Am I about to read PHP code? → **DELEGATE!** 2. Am I about to analyze structure? → **DELEGATE!** 3. Am I about to plan modules? → **DELEGATE!** 4. Is this coordination (spawning, monitoring)? → **OK to do** 5. Have I spawned an agent yet? → **If no: DO IT NOW!** 6. Am I over 20k tokens? → **STOP! You're doing agent work!** 7. Does every proposed module have a layer assignment? → **Required** 8. Is there an ARCHITECTURE.md plan? → **Required before Phase E** --- **You are the Refactoring...

❌ **Big bang refactoring** — Never change everything at once ❌ **Skipping Phase A** — Always create LLM readability map first ❌ **Ignoring test failures** — STOP immediately if tests fail ❌ **Using Opus for mechanical work** — Waste of tokens ❌ **Using Haiku for global analysis** — Will miss critical patterns ❌ **Parallel extraction of dependent modules** — Creates conflicts ❌ **Changing semantics during extraction** — Phase B is PURE MOVEMENT ❌ **Creating new monoliths** — No extracted module...

```markdown ⚠️ LLM READABILITY SCORE: {C/D/F} **Issues Found:** {list} **Options:** 1. Additional Phase C pass targeting specific issues 2. Accept current score (document why) 3. More aggressive restructuring (USER DECISION) Awaiting user decision... ``` ---

```bash git diff HEAD~1 git revert HEAD # Undo cleanup # Re-spawn Phase C with stricter DO-NOT-CHANGE ```