Require structured GenPage agent messages

evals/model-apps/genpage/EVAL_GUIDE.md

@@ -72,6 +72,9 @@ Layers 1 and 2 are automated. Layer 3 is human judgment by design — visual qua
 - Solution-selection question runs when (and only when) metadata work is needed
 - `check-auth.js` runs before `entity-builder` when entities need creating
 - `pac model genpage upload` invocations include `--prompt` with correct scoping
+- Every upload uses an escaped, structured `--agent-message` with OOB-style
+  Step-by-Step Agent Thoughts (bold numbered headings separated by blank lines),
+  a bulleted Summary, and Final Code sections
 - Prefix discipline holds across plan, entity-creation log, and resolved names
 
 **Runner:** `evals/model-apps/genpage/run-layer-1.js`
@@ -126,7 +129,7 @@ Plus per-eval `expectations` whose text starts with `Phase 5` (page-builder-spec
 
 All eval definitions live in `evals.json` alongside this file. The file contains:
 
-- `common_workflow_assertions`: 15 workflow checks every run must pass (prereqs, auth, solution selection gating, check-auth pre-flight, plan creation, workflow log, `--prompt` scoping, prefix discipline at plan-format / resolved-names / solution-alignment).
+- `common_workflow_assertions`: 16 workflow checks every run must pass (prereqs, auth, solution selection gating, check-auth pre-flight, plan creation, workflow log, `--prompt` scoping, structured `--agent-message` content, prefix discipline at plan-format / resolved-names / solution-alignment).
 - `common_code_assertions`: 18 code-quality checks the generated `.tsx` must pass (Fluent UI V9 only, no forbidden patterns, etc.).
 - `evals`: 16 test cases — each with `id`, `tier`, `prompt`, `data`, and per-eval `expectations`.
 

evals/model-apps/genpage/evals.json

@@ -14,6 +14,7 @@
     "Phase 1 (Planner): genpage-plan.md is written to the working directory, conforming to references/plan-schema.md",
     "Phase 2a: When entities need creating, scripts/check-auth.js runs and returns ok:true before entity-builder is invoked; on ok:false the orchestrator surfaces the message to the user and halts",
     "Phase 6 / 6.5 / 7.5 / Edit Phase 6 (--prompt scoping): The FIRST pac model genpage upload for each new page passes --prompt with the FULL page description (from plan's ## User Requirements). EVERY SUBSEQUENT upload of an existing page (--page-id, no --add-to-sitemap) passes --prompt scoped to ONLY the delta of changes in that upload — never the full original description",
+    "Every pac model genpage upload records --agent-message with escaped # Agent Thoughts, OOB-style Step-by-Step Processing with bold numbered headings on new lines, a bulleted # Summary, and # Final Code sections",
     "Prefix discipline — plan format: Every name in `## Entity Creation Required` (table headings, column Suffix values, choice column suffixes, relationship Lookup Suffix values) is a bare suffix matching `^[a-z][a-z0-9]+$`. No value contains an underscore or a prefix. The prefix lives only in `## Environment` → `Publisher Prefix:`.",
     "Prefix discipline — resolved names: For every operation in `entity-creation-log.md`, the Resolved Full Name starts with the `Publisher Prefix:` value from the plan's `## Environment` followed by `_` and the bare suffix from the plan (e.g., Publisher Prefix `crb2b` + suffix `playername` → `crb2b_playername`).",
     "Prefix discipline — solution alignment: If the env has a dominant non-system custom prefix (>=50% of custom tables AND >=3 such tables), the planner surfaces this in the solution question and ordered the option whose publisher prefix matches the dominant prefix FIRST as the recommended choice. If the user picks a different-prefix solution, the planner logged a one-line warning before proceeding."

evals/model-apps/genpage/fixtures/1-account-card-gallery/workflow-log.md

@@ -55,7 +55,7 @@
 - Post-write icon verification: grep `from "@fluentui/react-icons"` in page.tsx; verified `BuildingRegular`, `MailRegular`, `PhoneRegular`, `GlobeRegular` against verified-icons.txt — all present
 
 ## Phase 6 — Deployment
-- `pac model genpage upload --app-id 12345678-1234-1234-1234-123456789abc --code-file account-card-gallery/page.tsx --data-sources 'account' --prompt "Build a page showing Account records as a gallery of cards. Include name, website, email, phone number. Make the gallery scrollable and each card clickable to open the Account record." --model claude-sonnet --name "Accounts Gallery" --agent-message "Account card gallery with click-to-open" --add-to-sitemap`
+- `pac model genpage upload --app-id 12345678-1234-1234-1234-123456789abc --code-file account-card-gallery/page.tsx --data-sources 'account' --prompt "Build a page showing Account records as a gallery of cards. Include name, website, email, phone number. Make the gallery scrollable and each card clickable to open the Account record." --model claude-sonnet --name "Accounts Gallery" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Designed a responsive card gallery using verified Account columns and click-to-open navigation.\n# Summary\n- Created an Account card gallery with website, email, phone, and record navigation.\n# Final Code\n" --add-to-sitemap`
 - Upload succeeded; page-id = abc12345-...
 
 ## Phase 8 — Summary

evals/model-apps/genpage/fixtures/11-recruitment-multi-page/workflow-log.md

@@ -59,17 +59,17 @@
 - Cross-page navigation: candidate-list.tsx links to interview-schedule.tsx via PAGEREF_interview-schedule placeholder; interview-schedule.tsx links to hiring-metrics.tsx via PAGEREF_hiring-metrics
 
 ## Phase 6 — Deployment
-- pac model genpage upload --app-id 44444444-3333-4444-5555-666666666666 --code-file recruitment-pages/candidate-list.tsx --data-sources 'contact,appointment' --prompt "Candidate list page (eval 11)" --model claude-sonnet --name "Candidates" --agent-message "Candidate list" --add-to-sitemap
-- pac model genpage upload --app-id 44444444-3333-4444-5555-666666666666 --code-file recruitment-pages/interview-schedule.tsx --data-sources 'contact,appointment' --prompt "Interview schedule page (eval 11)" --model claude-sonnet --name "Schedule" --agent-message "Interview schedule" --add-to-sitemap
-- pac model genpage upload --app-id 44444444-3333-4444-5555-666666666666 --code-file recruitment-pages/hiring-metrics.tsx --data-sources 'contact,appointment' --prompt "Hiring metrics dashboard (eval 11)" --model claude-sonnet --name "Metrics" --agent-message "Hiring metrics" --add-to-sitemap
+- pac model genpage upload --app-id 44444444-3333-4444-5555-666666666666 --code-file recruitment-pages/candidate-list.tsx --data-sources 'contact,appointment' --prompt "Candidate list page (eval 11)" --model claude-sonnet --name "Candidates" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- Candidate list\n# Final Code\n" --add-to-sitemap
+- pac model genpage upload --app-id 44444444-3333-4444-5555-666666666666 --code-file recruitment-pages/interview-schedule.tsx --data-sources 'contact,appointment' --prompt "Interview schedule page (eval 11)" --model claude-sonnet --name "Schedule" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- Interview schedule\n# Final Code\n" --add-to-sitemap
+- pac model genpage upload --app-id 44444444-3333-4444-5555-666666666666 --code-file recruitment-pages/hiring-metrics.tsx --data-sources 'contact,appointment' --prompt "Hiring metrics dashboard (eval 11)" --model claude-sonnet --name "Metrics" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- Hiring metrics\n# Final Code\n" --add-to-sitemap
 
 ## Phase 6.5 — Cross-page reference resolution
 - Detected PAGEREF_interview-schedule and PAGEREF_hiring-metrics tokens in candidate-list.tsx and interview-schedule.tsx
 - Built filename→page-id map from Phase 6 returned IDs (sorted by length descending)
 - Replaced quoted `"PAGEREF_<name>"` tokens with the matching GUIDs (word-boundary safe)
 - Re-uploaded affected files with --page-id and a delta --prompt: "Resolve cross-page navigation placeholders to real page GUIDs"
-- pac model genpage upload --app-id 44444444-3333-4444-5555-666666666666 --code-file recruitment-pages/candidate-list.tsx --page-id <resolved-id> --data-sources 'contact,appointment' --prompt "Resolve cross-page navigation placeholders to real page GUIDs" --model claude-sonnet --agent-message "PAGEREF resolution"
-- pac model genpage upload --app-id 44444444-3333-4444-5555-666666666666 --code-file recruitment-pages/interview-schedule.tsx --page-id <resolved-id> --data-sources 'contact,appointment' --prompt "Resolve cross-page navigation placeholders to real page GUIDs" --model claude-sonnet --agent-message "PAGEREF resolution"
+- pac model genpage upload --app-id 44444444-3333-4444-5555-666666666666 --code-file recruitment-pages/candidate-list.tsx --page-id <resolved-id> --data-sources 'contact,appointment' --prompt "Resolve cross-page navigation placeholders to real page GUIDs" --model claude-sonnet --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- PAGEREF resolution\n# Final Code\n"
+- pac model genpage upload --app-id 44444444-3333-4444-5555-666666666666 --code-file recruitment-pages/interview-schedule.tsx --page-id <resolved-id> --data-sources 'contact,appointment' --prompt "Resolve cross-page navigation placeholders to real page GUIDs" --model claude-sonnet --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- PAGEREF resolution\n# Final Code\n"
 - No unresolved PAGEREF tokens remained
 
 ## Phase 8 — Summary

evals/model-apps/genpage/fixtures/11-recruitment-pages-real/workflow-log.md

@@ -103,9 +103,9 @@ User response: approved
 - Icons verified (all 6 files' imports confirmed in verified-icons.txt)
 
 ## Phase 6 — Deploy
-- Command 1: `pac model genpage upload --app-id e2666772-3c55-f111-a821-000d3a380330 --code-file D:/temp/recruitment-app/candidate-list.tsx --name "Candidate List" --data-sources "contact" --prompt "Build me three pages for a recruitment app: a candidate list page, an interview schedule page, and a hiring metrics dashboard. Use the contact and appointment entities." --model "claude-sonnet-4-6" --agent-message "Searchable filterable Contact list with caching, status badges, row-click navigation, pagination" --add-to-sitemap`
-- Command 2: `pac model genpage upload --app-id e2666772-3c55-f111-a821-000d3a380330 --code-file D:/temp/recruitment-app/interview-schedule.tsx --name "Interview Schedule" --data-sources "appointment,contact" --prompt "Build me three pages for a recruitment app: a candidate list page, an interview schedule page, and a hiring metrics dashboard. Use the contact and appointment entities." --model "claude-sonnet-4-6" --agent-message "Appointment DataGrid grouped by day, DatePicker filter, status filter, candidate name via OData annotation" --add-to-sitemap`
-- Command 3: `pac model genpage upload --app-id e2666772-3c55-f111-a821-000d3a380330 --code-file D:/temp/recruitment-app/hiring-metrics-dashboard.tsx --name "Hiring Metrics Dashboard" --data-sources "contact,appointment" --prompt "Build me three pages for a recruitment app: a candidate list page, an interview schedule page, and a hiring metrics dashboard. Use the contact and appointment entities." --model "claude-sonnet-4-6" --agent-message "KPI dashboard: 4 stat cards, D3 role bar chart, recent interviews list. PAGEREF_ placeholders resolved in Phase 6.5." --add-to-sitemap`
+- Command 1: `pac model genpage upload --app-id e2666772-3c55-f111-a821-000d3a380330 --code-file D:/temp/recruitment-app/candidate-list.tsx --name "Candidate List" --data-sources "contact" --prompt "Build me three pages for a recruitment app: a candidate list page, an interview schedule page, and a hiring metrics dashboard. Use the contact and appointment entities." --model "claude-sonnet-4-6" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- Searchable filterable Contact list with caching, status badges, row-click navigation, pagination\n# Final Code\n" --add-to-sitemap`
+- Command 2: `pac model genpage upload --app-id e2666772-3c55-f111-a821-000d3a380330 --code-file D:/temp/recruitment-app/interview-schedule.tsx --name "Interview Schedule" --data-sources "appointment,contact" --prompt "Build me three pages for a recruitment app: a candidate list page, an interview schedule page, and a hiring metrics dashboard. Use the contact and appointment entities." --model "claude-sonnet-4-6" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- Appointment DataGrid grouped by day, DatePicker filter, status filter, candidate name via OData annotation\n# Final Code\n" --add-to-sitemap`
+- Command 3: `pac model genpage upload --app-id e2666772-3c55-f111-a821-000d3a380330 --code-file D:/temp/recruitment-app/hiring-metrics-dashboard.tsx --name "Hiring Metrics Dashboard" --data-sources "contact,appointment" --prompt "Build me three pages for a recruitment app: a candidate list page, an interview schedule page, and a hiring metrics dashboard. Use the contact and appointment entities." --model "claude-sonnet-4-6" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- KPI dashboard: 4 stat cards, D3 role bar chart, recent interviews list. PAGEREF_ placeholders resolved in Phase 6.5.\n# Final Code\n" --add-to-sitemap`
 
 ## Phase 6 — Deploy (results)
 - Command 1 result: page-id = 492d8c42-b5fc-4ec5-ad44-6809c6673e9a, status = success
@@ -115,7 +115,7 @@ User response: approved
 ## Phase 6.5 — Navigation fix-up
 - PAGEREF_ map: candidate-list → 492d8c42-b5fc-4ec5-ad44-6809c6673e9a, interview-schedule → 07ad1d40-5ca8-494a-aeac-ff0548acb6c0
 - Replaced 2 tokens in hiring-metrics-dashboard.tsx (longer key first: interview-schedule, then candidate-list)
-- Re-upload command: `pac model genpage upload --app-id e2666772-3c55-f111-a821-000d3a380330 --page-id e2022b27-548b-43be-95bd-4dafa02e060c --code-file D:/temp/recruitment-app/hiring-metrics-dashboard.tsx --data-sources "contact,appointment" --prompt "Resolve cross-page navigation placeholders to real page GUIDs (post-deploy fix-up)" --model "claude-sonnet-4-6" --agent-message "Replaced PAGEREF_candidate-list with 492d8c42... and PAGEREF_interview-schedule with 07ad1d40..."`
+- Re-upload command: `pac model genpage upload --app-id e2666772-3c55-f111-a821-000d3a380330 --page-id e2022b27-548b-43be-95bd-4dafa02e060c --code-file D:/temp/recruitment-app/hiring-metrics-dashboard.tsx --data-sources "contact,appointment" --prompt "Resolve cross-page navigation placeholders to real page GUIDs (post-deploy fix-up)" --model "claude-sonnet-4-6" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- Replaced PAGEREF_candidate-list with 492d8c42... and PAGEREF_interview-schedule with 07ad1d40...\n# Final Code\n"`
 - Result: page-id = e2022b27-548b-43be-95bd-4dafa02e060c, status = success
 
 ## Phase 7 — Browser verification

evals/model-apps/genpage/fixtures/13-contact-localization/workflow-log.md

@@ -64,7 +64,7 @@
 - Post-write icon verification: PeopleRegular verified against verified-icons.txt
 
 ## Phase 6 — Deployment
-- pac model genpage upload --app-id 55555555-4444-5555-6666-777777777777 --code-file contact-localized/page.tsx --data-sources 'contact' --prompt "Build a page showing Contact records with name, email, and phone." --model claude-sonnet --name "Contacts (Localized)" --agent-message "Contact list with en-US/ar-SA/fr-FR translations and RTL support" --add-to-sitemap
+- pac model genpage upload --app-id 55555555-4444-5555-6666-777777777777 --code-file contact-localized/page.tsx --data-sources 'contact' --prompt "Build a page showing Contact records with name, email, and phone." --model claude-sonnet --name "Contacts (Localized)" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- Contact list with en-US/ar-SA/fr-FR translations and RTL support\n# Final Code\n" --add-to-sitemap
 
 ## Phase 8 — Summary
 - 1 page deployed: page.tsx → "Contacts (Localized)" in Sales Hub

evals/model-apps/genpage/fixtures/15-support-tickets-real/workflow-log.md

@@ -150,7 +150,7 @@ genpage-plan.md written to D:/temp/support-tickets/genpage-plan.md
   - ArrowClockwiseRegular ✓
 
 ## Phase 6 — Deploy
-- Command: `pac model genpage upload --app-id 3fc905b9-7854-f111-a821-70a8a59ce7bc --code-file D:/temp/support-tickets/support-tickets.tsx --name "Support Tickets" --data-sources "cr_ticket" --prompt "Build a page tracking support tickets. I need a new cr_ticket entity with a priority choice column (Low, Medium, High, Critical), a status choice column (Open, In Progress, Resolved, Closed), and a due date column." --model "claude-sonnet-4-6" --agent-message "Filterable DataGrid of cr_ticket records: search by name, dropdown filters for status and priority, controlled column sort defaulting to due date asc, color-coded priority badges (Low=green, Medium=warning, High=severe, Critical=danger), status pills, overdue indicator on past-due active tickets, row click opens Dataverse record form, window cache, refresh button." --add-to-sitemap`
+- Command: `pac model genpage upload --app-id 3fc905b9-7854-f111-a821-70a8a59ce7bc --code-file D:/temp/support-tickets/support-tickets.tsx --name "Support Tickets" --data-sources "cr_ticket" --prompt "Build a page tracking support tickets. I need a new cr_ticket entity with a priority choice column (Low, Medium, High, Critical), a status choice column (Open, In Progress, Resolved, Closed), and a due date column." --model "claude-sonnet-4-6" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- Filterable DataGrid of cr_ticket records: search by name, dropdown filters for status and priority, controlled column sort defaulting to due date asc, color-coded priority badges (Low=green, Medium=warning, High=severe, Critical=danger), status pills, overdue indicator on past-due active tickets, row click opens Dataverse record form, window cache, refresh button.\n# Final Code\n" --add-to-sitemap`
 - Result: page-id = e460d8cf-7ca7-4745-a895-28f1425fa377, status = success
 
 ## Phase 6.5 — Navigation fix-up

evals/model-apps/genpage/fixtures/17-weather-connectors-off/workflow-log.md

@@ -61,7 +61,7 @@
 - Post-write icon verification: grep `from "@fluentui/react-icons"` in weather-dashboard.tsx; verified `TemperatureRegular`, `WeatherSunnyRegular`, `WaterRegular` against verified-icons.txt — all present
 
 ## Phase 6 — Deployment
-- `pac model genpage upload --app-id aa112233-1122-1122-1122-aabbccdd1234 --code-file seattle-weather-dashboard/weather-dashboard.tsx --prompt "Build a dashboard showing the current weather for Seattle with temperature, conditions, and humidity." --model claude-sonnet --name "Seattle Weather" --agent-message "Weather dashboard with temperature, conditions, and humidity" --add-to-sitemap`
+- `pac model genpage upload --app-id aa112233-1122-1122-1122-aabbccdd1234 --code-file seattle-weather-dashboard/weather-dashboard.tsx --prompt "Build a dashboard showing the current weather for Seattle with temperature, conditions, and humidity." --model claude-sonnet --name "Seattle Weather" --agent-message "# Agent Thoughts\nStep-by-Step Processing\n\n**1. Requirements Analysis**\n- Applied the requested page changes using verified requirements.\n# Summary\n- Weather dashboard with temperature, conditions, and humidity\n# Final Code\n" --add-to-sitemap`
 - (--data-sources omitted — mock-data page)
 - Upload succeeded