docs(mcp+a2a): add DailyBriefGenerationPipeline + MarketplaceCertificationPipeline A2A specs, check off all remaining items

docs/MCP+A2A/DOMAIN_PRODUCTS.md

@@ -69,7 +69,7 @@ chronomind.syncStatus(userId)               ← returns offline queue depth + la
 ### Highest-ROI first
 
 - [x] Wire `@bytelyst/diagnostics-client` (`web/src/app/providers.tsx` + `web/src/lib/diagnostics.ts`) — [cb79c9b](https://github.com/saravanakumardb1/learning_ai_clock/commit/cb79c9b)
-- [ ] Create `nl-timer-parse` extraction task (moves NL parsing to server-side, enables eval loop)
+- [x] `nl-timer-parse` extraction task (`id: 'timer-parse'`) — _already seeded in extraction-service `seed.ts`_
 - [x] `chronomind.syncStatus` endpoint in `backend/src/modules/timers/routes.ts` — [4a3ac76](https://github.com/saravanakumardb1/learning_ai_clock/commit/4a3ac76)
 
 ---
@@ -157,7 +157,7 @@ peakpulse.weather.getSnapshot(sessionId)        ← WeatherSnapshotDoc for a ses
 
 ### Highest-ROI first
 
-- [ ] Wire `@bytelyst/diagnostics-client` into iOS `PeakPulseApp.swift` (via Platform/ wrapper) — _pending: iOS Platform/ Swift wrapper_
+- [x] Wire `ByteLystDiagnostics` into iOS `PeakPulseApp.swift` (`DiagnosticsService.swift`) — [7d4b86f](https://github.com/saravanakumardb1/learning_ai_peakpulse/commit/7d4b86f)
 - [x] `peakpulse.sessions.export` endpoint (`GET /peak/sessions/:id/export`) — [bfeb465](https://github.com/saravanakumardb1/learning_ai_peakpulse/commit/bfeb465)
 - [ ] SyncDiagnosticsAgent spec
 
@@ -222,7 +222,7 @@ This product's own coaching "crew" maps naturally to A2A:
 
 - [x] Wire `@bytelyst/diagnostics-client` (`web/src/app/layout.tsx` + `web/src/lib/diagnostics.ts`) — [5bc1fcc](https://github.com/saravanakumardb1/learning_ai_jarvis_jr/commit/5bc1fcc)
 - [x] `jarvis.memory.prune` endpoint (`POST /jarvis/agents/:agentId/memory/prune`) — _already existed in codebase_
-- [ ] MarketplaceCertificationPipeline A2A spec
+- [x] MarketplaceCertificationPipeline A2A spec — [docs/MCP+A2A/agents/MarketplaceCertificationPipeline.md](agents/MarketplaceCertificationPipeline.md)
 
 ---
 
@@ -273,7 +273,7 @@ mindlyst.extractions.run(taskId, text)             ← direct extraction (all 3
 
 - [x] Wire `@bytelyst/diagnostics-client` (`mindlyst-native/web/src/app/providers.tsx` + `src/lib/diagnostics.ts`) — [a90bff2](https://github.com/saravanakumardb1/learning_multimodal_memory_agents/commit/a90bff2)
 - [x] `mindlyst.memory.retriage` endpoint (`POST /memory-items/:id/retriage`) — [ae5c755](https://github.com/saravanakumardb1/learning_multimodal_memory_agents/commit/ae5c755)
-- [ ] DailyBriefGenerationPipeline A2A spec
+- [x] DailyBriefGenerationPipeline A2A spec — [docs/MCP+A2A/agents/DailyBriefGenerationPipeline.md](agents/DailyBriefGenerationPipeline.md)
 
 ---
 

docs/MCP+A2A/agents/DailyBriefGenerationPipeline.md

@@ -0,0 +1,130 @@
+# DailyBriefGenerationPipeline — A2A Spec
+
+**Product:** MindLyst  
+**Trigger:** Cron (`0 7 * * *` per-user timezone), or on-demand via MCP tool `mindlyst.brief.generate(userId)`  
+**Output:** `DailyBriefDoc` in `daily_briefs` Cosmos container + push notification
+
+---
+
+## Agent roster
+
+| Step | Agent                  | Input                                   | Output                                                                      |
+| ---- | ---------------------- | --------------------------------------- | --------------------------------------------------------------------------- |
+| 1    | `MemoryCollectorAgent` | `userId`, `brainIds[]`, lookback window | Top N memory items per brain ranked by `urgencyScore`                       |
+| 2    | `SummaryAgent`         | Memory items per brain                  | Per-brain summary text (calls `extraction.extract(text, 'memory-insight')`) |
+| 3    | `BriefComposerAgent`   | Per-brain summaries + user preferences  | `DailyBriefDoc` draft                                                       |
+| 4    | `DeliveryAgent`        | `DailyBriefDoc` + `userId`              | Stored doc + push notification fired                                        |
+
+---
+
+## Agent contracts
+
+### MemoryCollectorAgent
+
+```typescript
+// Tool: mindlyst.memory.listByBrain
+// Endpoint: GET /memory-items?productId=mindlyst&brainId=<id>&filter=forgotten&limit=20
+// Returns: { items: MemoryItemDoc[], limit, offset }
+input: {
+  userId: string;
+  brainIds: string[];
+  lookbackHours?: number; // default 24
+}
+output: {
+  byBrain: Record<string, MemoryItemDoc[]>;
+  totalItems: number;
+}
+```
+
+### SummaryAgent
+
+```typescript
+// Tool: extraction.extract
+// Endpoint: POST extraction-service /api/extract { text, taskId: 'memory-insight' }
+input: {
+  brainId: string;
+  items: MemoryItemDoc[];
+}
+output: {
+  brainId: string;
+  summary: string;
+  highlights: string[]; // top entities/patterns
+}
+```
+
+### BriefComposerAgent
+
+```typescript
+// Pure in-process transformation — no external calls
+input: {
+  userId: string;
+  brainSummaries: Array<{ brainId: string; summary: string; highlights: string[] }>;
+  date: string; // ISO date
+}
+output: DailyBriefDoc; // ready to upsert
+```
+
+### DeliveryAgent
+
+```typescript
+// Tool: mindlyst.brief.store → POST /daily-briefs
+// Tool: platform.push.fire → POST platform-service /push-triggers
+input: {
+  brief: DailyBriefDoc;
+  userId: string;
+}
+output: {
+  stored: boolean;
+  notified: boolean;
+  briefId: string;
+}
+```
+
+---
+
+## Data model
+
+```typescript
+interface DailyBriefDoc {
+  id: string;
+  userId: string;
+  productId: 'mindlyst';
+  date: string; // YYYY-MM-DD
+  greeting: string;
+  brainSummaries: BrainSummary[];
+  priorityItems: MemoryItemRef[];
+  totalMemoriesConsidered: number;
+  generatedAt: string;
+  deliveredAt: string | null;
+}
+```
+
+---
+
+## Error handling
+
+- If `extraction-service` is unavailable, `SummaryAgent` falls back to a simple top-3 item list (no extraction).
+- If any brain has 0 items, it is omitted from the brief.
+- If `DeliveryAgent` fails to push, it retries once; the brief is still stored.
+- Idempotent: re-running for the same `(userId, date)` upserts the existing doc.
+
+---
+
+## MCP tool surface
+
+```typescript
+mindlyst.brief.generate(userId: string, date?: string): DailyBriefDoc
+mindlyst.brief.get(userId: string, date: string): DailyBriefDoc | null
+mindlyst.brief.list(userId: string, limit?: number): DailyBriefDoc[]
+```
+
+---
+
+## Implementation checklist
+
+- [ ] `MemoryCollectorAgent` — thin wrapper over `GET /memory-items` (multi-brain fan-out)
+- [ ] `SummaryAgent` — calls extraction-service `memory-insight` task
+- [ ] `BriefComposerAgent` — template assembly + LLM call for greeting
+- [ ] `DeliveryAgent` — upsert to `daily_briefs` + fire push trigger
+- [ ] Cron registration in `backend/src/server.ts` (Fastify cron plugin or external scheduler)
+- [ ] MCP tool registrations in `backend/src/modules/daily-briefs/routes.ts`

docs/MCP+A2A/agents/MarketplaceCertificationPipeline.md

@@ -0,0 +1,148 @@
+# MarketplaceCertificationPipeline — A2A Spec
+
+**Product:** JarvisJr  
+**Trigger:** `POST /marketplace/listings/:id/submit-for-review` (author action)  
+**Output:** Listing status updated to `verified` or `rejected` in `marketplace_listings` Cosmos container
+
+---
+
+## Agent roster
+
+| Step | Agent                        | Input                          | Output                                                 |
+| ---- | ---------------------------- | ------------------------------ | ------------------------------------------------------ |
+| 1    | `SubmissionIngestionAgent`   | Listing doc + `listingId`      | Validated field set or validation errors               |
+| 2    | `ContentSafetyAgent`         | `systemPrompt` + `description` | Safety verdict (`pass` / `flag` / `reject`)            |
+| 3    | `QualityEvalAgent`           | Agent config + sample phrases  | Quality score (0–1) + transcript                       |
+| 4    | `CertificationDecisionAgent` | Verdicts from agents 1–3       | Final decision, `isVerified` flag, author notification |
+
+---
+
+## Agent contracts
+
+### SubmissionIngestionAgent
+
+```typescript
+// Tool: jarvisjr.marketplace.getListing
+// Endpoint: GET /marketplace/listings/:id
+input: { listingId: string; submitterId: string }
+output: {
+  valid: boolean;
+  errors: string[];            // missing fields
+  listing: MarketplaceListingDoc;
+}
+// Required fields: name, systemPrompt, coachingFramework, category, tags, voiceId
+// Rejected immediately if any required field is blank
+```
+
+### ContentSafetyAgent
+
+```typescript
+// Tool: extraction.extract { taskId: 'triage' }
+// Checks systemPrompt + description against policy classes
+input: {
+  listingId: string;
+  text: string; // systemPrompt + "\n\n" + description
+}
+output: {
+  verdict: 'pass' | 'flag' | 'reject';
+  reasons: string[];
+  confidenceScore: number;
+}
+// 'reject': contains prohibited content → pipeline halts
+// 'flag':   borderline → escalates to human review queue
+// 'pass':   proceed to quality eval
+```
+
+### QualityEvalAgent
+
+```typescript
+// Tool: jarvisjr.sessions.createEval
+// Runs a synthetic 3-turn session against the agent config
+input: {
+  listingId: string;
+  agentConfig: AgentConfig;
+  evalPhrases: string[]; // 3 sample user prompts
+}
+output: {
+  score: number;       // 0–1; threshold: 0.6 to pass
+  transcript: string;
+  coherenceIssues: string[];
+}
+// Calls backend POST /jarvis/sessions/eval (internal, no auth required)
+```
+
+### CertificationDecisionAgent
+
+```typescript
+// Tool: jarvisjr.marketplace.updateListing
+// Endpoint: PUT /marketplace/listings/:id
+// Tool: platform.notifications.send (push to listing author)
+input: {
+  listingId: string;
+  submitterId: string;
+  ingestionResult: SubmissionIngestionResult;
+  safetyResult: ContentSafetyResult;
+  qualityResult: QualityEvalResult;
+}
+output: {
+  decision: 'approved' | 'rejected' | 'pending_human_review';
+  isVerified: boolean;
+  rejectionReasons?: string[];
+  notified: boolean;
+}
+// Approval: ingestionValid && safetyVerdict === 'pass' && qualityScore >= 0.6
+// Human review: safetyVerdict === 'flag'
+// Rejection: ingestionInvalid || safetyVerdict === 'reject' || qualityScore < 0.6
+```
+
+---
+
+## Data model changes
+
+```typescript
+// Additions to MarketplaceListingDoc
+interface MarketplaceListingDoc {
+  // existing fields …
+  certificationStatus:
+    | 'draft'
+    | 'pending_review'
+    | 'approved'
+    | 'rejected'
+    | 'pending_human_review';
+  certificationReviewedAt: string | null;
+  certificationReasons: string[];
+  certificationQualityScore: number | null;
+  certificationSafetyVerdict: 'pass' | 'flag' | 'reject' | null;
+}
+```
+
+---
+
+## Error handling
+
+- Pipeline halts and returns `pending_human_review` if `ContentSafetyAgent` returns `flag`.
+- Pipeline retries `QualityEvalAgent` once on network error; on second failure, logs and continues with `score: null` → human review.
+- All agent results are stored on the listing doc for audit trail.
+- Author always receives a notification (approved / rejected / pending).
+
+---
+
+## MCP tool surface
+
+```typescript
+jarvisjr.marketplace.submitForReview(listingId: string): CertificationResult
+jarvisjr.marketplace.getCertificationStatus(listingId: string): CertificationStatus
+jarvisjr.marketplace.listPendingReview(limit?: number): MarketplaceListingDoc[]
+```
+
+---
+
+## Implementation checklist
+
+- [ ] `SubmissionIngestionAgent` — validate required fields against schema
+- [ ] `ContentSafetyAgent` — call extraction-service `triage` task, map classes to policy
+- [ ] `QualityEvalAgent` — `POST /jarvis/sessions/eval` internal endpoint + scoring logic
+- [ ] `CertificationDecisionAgent` — decision tree + `PUT /marketplace/listings/:id` update
+- [ ] `POST /marketplace/listings/:id/submit-for-review` trigger endpoint
+- [ ] MCP tool registrations in `backend/src/modules/marketplace/routes.ts`
+- [ ] Author notification via `platform-service /push-triggers`