@polymorphism-tech/morph-spec 8.33.0 → 8.35.0

package/framework/standards/ai-agents/workflows.md

@@ -5,7 +5,7 @@
 > **Keywords:** workflow, orchestration, fan-in, fan-out, multi-agent workflow, sequential workflow
 > **Read by Claude in:** plan (only when a genuine fan-in/out or long-running need appears)
 
-**Verified against:** MAF 1.0 GA. Last-verified: 2026-05-19.
+**Verified against:** Microsoft.Agents.AI 1.20.0 (ai-pin 2026-09-08) + Microsoft.Agents.AI.Workflows 1.20.0, Microsoft.Agents.AI.Workflows.Declarative 1.20.0 e Microsoft.Agents.AI.Declarative `1.20.0-rc1` (**os três fora do pin**; versões apuradas no nuget.org em 2026-09-08). **Verificação documental**, sem cláusula `provado por` — o ai-kit não monta workflows. **A localização dos tipos de workflow foi medida por reflexão em 2026-09-08 e CORRIGE este standard**: eles não estão no pacote base. Last-verified: 2026-09-08.
 
 ---
 
@@ -43,19 +43,39 @@ Para o critério de HITL/long-running acima, a implementação (`RequestPort`, `
 
 ## Packages
 
-Workflow APIs (`AgentWorkflowBuilder`, `WorkflowBuilder`, `InProcessExecution`) estão no pacote GA principal:
+> **Correção de 2026-09-08, medida.** A versão anterior deste standard dizia que `AgentWorkflowBuilder`, `WorkflowBuilder` e `InProcessExecution` estavam no pacote base `Microsoft.Agents.AI`. **Não estão.** Varredura dos tipos exportados de `Microsoft.Agents.AI` 1.20.0: **zero** tipos de workflow. Todos os 132 tipos de workflow — `AgentWorkflowBuilder`, `WorkflowBuilder`, `InProcessExecution`, `HandoffWorkflowBuilder`, `GroupChatWorkflowBuilder`, `MagenticWorkflowBuilder`, `CheckpointManager` — vivem em **`Microsoft.Agents.AI.Workflows`**, um pacote **separado**. Seguir o texto antigo dá erro de compilação, não aviso.
 
 ```xml
-<PackageReference Include="Microsoft.Agents.AI" Version="1.0.*" />
+<!-- Base MAF — no ai-pin.json -->
+<PackageReference Include="Microsoft.Agents.AI" Version="1.20.0" />
+
+<!-- Workflows: PACOTE SEPARADO. Estável. Fora do pin; apurado no nuget.org em 2026-09-08. -->
+<PackageReference Include="Microsoft.Agents.AI.Workflows" Version="1.20.0" />
 ```
 
-Para DI helpers (`AddAIAgent`):
+Para DI helpers (`AddAIAgent`, `AddSequentialWorkflow`):
 
 ```xml
-<PackageReference Include="Microsoft.Agents.AI.Hosting" Version="1.0.*" />
+<!-- PRERELEASE: não há versão estável. Fora do pin; apurado em 2026-09-08. -->
+<PackageReference Include="Microsoft.Agents.AI.Hosting" Version="1.20.0-preview.260831.1" />
 ```
 
-> Sem pacotes "preview" separados. Sem `Microsoft.Extensions.AI 9.x`. As APIs de Workflow são parte do `Microsoft.Agents.AI` 1.0 GA.
+### Declarativo: workflow é estável, agente ainda não
+
+Os dois costumam ser citados juntos e têm maturidade **diferente**. Apurado no nuget.org em 2026-09-08:
+
+| Pacote | Última versão | Estável? |
+|---|---|---|
+| `Microsoft.Agents.AI.Workflows.Declarative` (workflows em YAML) | **1.20.0** | **Sim** — estáveis desde a 1.9 |
+| `Microsoft.Agents.AI.Declarative` (agentes declarativos) | `1.20.0-rc1` | **Não** — ainda exige `--prerelease` |
+
+Ou seja: descrever o **grafo** em YAML é caminho estável; descrever o **agente** em YAML ainda não.
+
+### Mudança de comportamento na 1.17
+
+A partir da **1.17**, um workflow **falha quando um agente participante retorna erro** — antes o erro podia ser absorvido e o grafo seguia com um resultado parcial. Consequência: um workflow que "sempre funcionou" pode passar a falhar depois do upgrade, e isso é o comportamento **correto**. Trate erro de agente explicitamente (branch de erro no grafo, ou tratamento dentro do executor) em vez de contar com a absorção antiga.
+
+> Sem `Microsoft.Extensions.AI 9.x`. Nenhuma faixa flutuante — SDK de cadência semanal com `1.0.*` é o defeito que `ai-agents-setup` §1 proíbe.
 
 ---
 
@@ -187,4 +207,4 @@ Sempre use `await using` no `StreamingRun` / `Run` — é `IAsyncDisposable`.
 
 ---
 
-*MORPH-SPEC by Polymorphism Tech — ai-agents/workflows.md v2.0 (2026-05-19)*
+*MORPH-SPEC by Polymorphism Tech — ai-agents/workflows.md v2.0 (2026-09-08)*

package/framework/standards/frontend/design-system/ai-image-generation.md

@@ -288,4 +288,16 @@ If none are good, regenerate at the same step with the prompt adjusted by the fe
 - [ ] Images headed for video framed without faces
 - [ ] Favicon and 1200x630 OG image produced
 
+## Boundary: Runtime Generation Is Another Standard
+
+This standard is **authoring time**: a designer, an MCP tool, and assets that end up committed in the
+repository. When the **product itself** generates images at runtime — a .NET backend calling a paid
+API per request, with cost, review and provenance — the standard is
+`ai-agents/modalities-image-gen.md`, and the surrounding job/review/labelling pipeline is
+`ai-agents/media-pipeline.md`.
+
+The two do not overlap: nothing here talks about paid calls per user, and nothing there repeats the
+brand prompt craft above. The rule that never crosses the boundary is the hard one from
+`ai-agents/media-pipeline.md`: **never generate media inside a conversation turn**.
+
 *MORPH-SPEC by Polymorphism Tech*

package/framework/standards/frontend/design-system/ai-video-generation.md

@@ -110,4 +110,17 @@ TECHNICAL: img2vid with image (start) + endImage (end). No audio. No watermark.
 - [ ] Draft at 480p/4s approved before any 1080p render
 - [ ] Frames re-extracted with a bumped cache-bust after the final render
 
+## Boundary: Runtime Generation Is Another Standard
+
+This standard is **authoring time**: a human driving the seedance MCP in a session, footage that ends
+up committed as a page asset. When the **product itself** generates video at runtime — a .NET worker
+creating a paid task, polling an operation, downloading bytes into its own blob store — the standard
+is `ai-agents/media-video.md` (Veo via `Google.GenAI`, the HTTP-only providers, cost per second, the
+2-day retention), and the surrounding job/review/labelling pipeline is
+`ai-agents/media-pipeline.md`.
+
+The face-moderation workaround in Section 3 above is the piece `ai-agents/media-video.md` cites
+rather than repeats. The rule that never crosses the boundary is the hard one from
+`ai-agents/media-pipeline.md`: **never generate media inside a conversation turn**.
+
 *MORPH-SPEC by Polymorphism Tech*

package/framework/standards/infrastructure/docker/local-compose-isolation.md

@@ -0,0 +1,90 @@
+# Local Compose Isolation Standard
+
+> **Scope:** any stack that runs `docker compose` locally (Next.js + Neon, .NET)
+> **Layer:** 2 (on keyword)
+> **Keywords:** docker compose, worktree, ports, container_name, POSTGRES_PORT, COMPOSE_PROJECT_NAME, local dev, e2e stack
+> **Load When:** a compose file, a dev-server port, or a worktree stack is being written or debugged
+
+**Verified against:** Docker Compose v2 + morph-spec 8.34 (dedicated stack per worktree). Last-verified: 2026-09-08.
+
+---
+
+## Why this exists
+
+Every feature worktree runs its **own** compose stack — its own project, its own postgres, its own volumes. That only holds if the compose file lets the harness pick the host ports and the project name. A file that hardcodes either one looks isolated and collides the moment a second worktree starts: `container_name` is never prefixed by the project, and a fixed `5432:5432` is one host port for everybody.
+
+The harness gives each worktree a **block of 10 ports** (`.morph/worktree.env`, written by `morph-spec worktree setup|provision`, regenerated by `worktree link --all`). The same values reach `docker compose` through the `verify` e2e node, `morph-spec e2e up`, the `PreToolUse:Bash` hook (which prefixes `npm run dev` / `dotnet run` / `docker compose` with them) and `morph-spec env` for a plain shell.
+
+## Port layout (one contract, everywhere)
+
+| Offset | Variable | Compose mapping |
+|---|---|---|
+| +0 | `PORT` | `"${PORT:-3000}:3000"` — web / dev server |
+| +1 | `API_PORT` (and `ASPNETCORE_URLS=http://localhost:<+1>` on .NET) | `"${API_PORT:-5000}:8080"` |
+| +2 | `POSTGRES_PORT` | `"${POSTGRES_PORT:-5432}:5432"` |
+| +3 | `REDIS_PORT` | `"${REDIS_PORT:-6379}:6379"` (reserved) |
+| +4..+9 | free | project-specific services |
+
+Also exported: `COMPOSE_PROJECT_NAME=morph-{feature}` (isolates container, network and volume names), `MORPH_PORT_BASE`, `MORPH_PORT_RANGE`, `PLAYWRIGHT_BASE_URL=http://localhost:<+0>`.
+
+## Core Rules
+
+- NEVER set `container_name` — Compose already names containers `<project>-<service>-1`, and an explicit name is shared by every worktree
+- ALWAYS publish host ports through a variable with a default: `"${PORT:-3000}:3000"`, never `"3000:3000"` nor `published: 3000`
+- ALWAYS keep the database name stable (`POSTGRES_DB: ${POSTGRES_DB:-app}`) — the project name already namespaces the container and the volume; baking a feature name into the DB name outlives the feature that scaffolded it
+- ALWAYS declare a healthcheck on every service — `docker compose up --wait` (what the e2e node runs) is meaningless without one
+- ALWAYS give `DATABASE_URL` an in-network default so the stack comes up with no `.env`: `postgresql://postgres:postgres@postgres:5432/${POSTGRES_DB:-app}`
+- ALWAYS read the base URL from the environment in `playwright.config.ts`: `process.env.PLAYWRIGHT_BASE_URL ?? \`http://localhost:${process.env.PORT ?? 3000}\``
+- NEVER put a cloud connection string (Neon branch etc.) in `.morph/worktree.env` — it is regenerated; credentials go in the gitignored file named by `config.json → e2e.auth.envFile`
+- NEVER put a variable from the port layout table above (nor `PLAYWRIGHT_BASE_URL`) in that credentials file — the harness **ignores** it and says which key it ignored on stderr. Ports come from the allocated block; the file carries credentials. A `.env.e2e` copied from a sibling worktree carries THAT worktree's block, and before the filter existed it silently repointed this stack at the neighbour's postgres
+
+## What the harness enforces
+
+- Inside a secondary worktree, a compose with a fixed host port or a `container_name` **fails the `verify` e2e node** (`compose-not-isolable`) — Gate 3 cannot pass on a stack that cannot run beside its siblings. At the primary root the same finding is a warning.
+- `morph-spec finish` tears the stack down **with volumes** (`--keep-volumes` opts out) and deletes the freshness stamp; `morph-spec worktree remove` tears it down without volumes (the feature is still open).
+- Two worktrees never share `POSTGRES_PORT`: the block is allocated first-free-index per feature and released only when the worktree is removed.
+- The credentials env file is read **after** the block but cannot outrank it. All four surfaces (`e2e up`, the `verify` e2e node, the teardown, `morph-spec env`) build their environment through the same `buildComposeEnv`, which drops the layout keys coming from that file and names them on **stderr** — stdout stays eval-safe for `eval "$(morph-spec env)"`. With no block allocated (outside a worktree) nothing is filtered: there the file is the only port source.
+
+## Minimal compose skeleton
+
+```yaml
+services:
+  app:
+    build: { context: ., dockerfile: Dockerfile.dev }
+    ports:
+      - "${PORT:-3000}:3000"
+    environment:
+      - DATABASE_URL=${DATABASE_URL:-postgresql://postgres:postgres@postgres:5432/${POSTGRES_DB:-app}}
+    depends_on:
+      postgres: { condition: service_healthy }
+    healthcheck:
+      test: ["CMD-SHELL", "node -e \"fetch('http://localhost:3000/').then(r=>process.exit(r.ok?0:1)).catch(()=>process.exit(1))\""]
+      interval: 10s
+      timeout: 5s
+      retries: 12
+  postgres:
+    image: postgres:15-alpine
+    ports:
+      - "${POSTGRES_PORT:-5432}:5432"
+    environment:
+      POSTGRES_USER: postgres
+      POSTGRES_PASSWORD: postgres
+      POSTGRES_DB: ${POSTGRES_DB:-app}
+    volumes:
+      - postgres-data:/var/lib/postgresql/data
+    healthcheck:
+      test: ["CMD-SHELL", "pg_isready -U postgres -d ${POSTGRES_DB:-app}"]
+      interval: 5s
+      timeout: 5s
+      retries: 10
+volumes:
+  postgres-data:
+```
+
+## Running it by hand
+
+```bash
+# inside the worktree, outside Claude Code
+eval "$(morph-spec env)"        # PowerShell: morph-spec env --powershell | Invoke-Expression
+docker compose up -d --wait     # lands in project morph-{feature}, on the block's ports
+```

package/framework/templates/REGISTRY.json

@@ -4,6 +4,11 @@
   "description": "Template Registry for MORPH-SPEC Framework - Technology-based organization for multi-stack reusability",
   "lastUpdated": "2026-02-27",
   "migration": "v1 → v2: Migrated all templates to framework/, organized by technology (code/dotnet/, infrastructure/azure/, etc.) instead of by stack. This enables future stacks (vue-firebase, angular-mongodb) to reuse existing technology templates.",
+  "kinds": {
+    "template": "Rendered to an output file. Rich shape: id, name, path, category, phase, outputName (+ engine, placeholders).",
+    "fragment": "A reference snippet read as an example, never rendered. Minimal shape: id, path, category, stack, description.",
+    "project": "A whole compilable project registered as ONE unit — its files are not consumable one by one. Minimal shape: id, path, category, stack, description; `path` ends in \"/\" because it is a DIRECTORY. Every tracked file under it counts as registered (validate-framework.js Check 12), and a project directory with no tracked file is an error."
+  },
   "categories": [
     "documentation",
     "code",
@@ -468,7 +473,7 @@
       "category": "infrastructure",
       "phase": "implement",
       "required": false,
-      "description": "Docker Compose for local development environment",
+      "description": "Docker Compose for local development — one isolated stack per worktree (ports from the block, no container_name)",
       "outputName": "docker-compose.yml",
       "placeholders": [],
       "stackSpecific": false,
@@ -1243,6 +1248,14 @@
       "stack": "dotnet",
       "description": "Typed MAF tool retrieving relevant chunks from the pgvector knowledge base"
     },
+    {
+      "id": "ai-kit",
+      "kind": "project",
+      "path": "dotnet/ai-kit/",
+      "category": "ai-agents",
+      "stack": "dotnet",
+      "description": "Morph.AiKit — solucao .NET compilavel que prova os standards ai-agents contra framework/ai-pin.json"
+    },
     {
       "id": "maf-model-registry-config",
       "kind": "fragment",

package/framework/templates/code/dotnet/ai-agents/PromptRepository.cs.template

@@ -1,95 +1,95 @@
-// TEMPLATE: PromptRepository.cs
-// PURPOSE: Loads agent system prompts from the agent_prompts table, with in-memory caching.
-// READS STANDARD: ai-agents-prompts-in-db
-// PLACEHOLDERS:
-//   {{PROJECT_NAMESPACE}}  — root namespace of the project
-// LAST-VERIFIED: 2026-05-19
-
-using Microsoft.EntityFrameworkCore;
-using Microsoft.Extensions.Caching.Memory;
-
-namespace {{PROJECT_NAMESPACE}}.AI;
-
-// ---------------------------------------------------------------------------
-// Entity — map this to the agent_prompts table (EF Core convention mapping).
-// Alternatively, configure via Fluent API in your DbContext.OnModelCreating.
-// ---------------------------------------------------------------------------
-
-/// <summary>
-/// EF Core entity for the agent_prompts table.
-/// Content is immutable once written — use INSERT + flip is_active for versioning.
-/// </summary>
-public sealed class AgentPrompt
-{
-    public Guid Id { get; init; } = Guid.CreateVersion7();
-    public required string AgentKey { get; init; }
-    public int Version { get; init; }
-    public required string Content { get; init; }
-    public bool IsActive { get; set; }
-    public DateTimeOffset CreatedAt { get; init; } = DateTimeOffset.UtcNow;
-}
-
-// ---------------------------------------------------------------------------
-// Interface
-// ---------------------------------------------------------------------------
-
-/// <summary>
-/// Provides the active system-prompt content for a named agent.
-/// </summary>
-public interface IPromptRepository
-{
-    /// <summary>
-    /// Returns the content of the currently active prompt for <paramref name="agentKey"/>.
-    /// </summary>
-    /// <exception cref="InvalidOperationException">
-    /// Thrown when no row with <c>is_active = true</c> exists for <paramref name="agentKey"/>.
-    /// </exception>
-    Task<string> GetActivePromptAsync(string agentKey, CancellationToken ct = default);
-}
-
-// ---------------------------------------------------------------------------
-// Implementation
-// ---------------------------------------------------------------------------
-
-// NOTE: Register in Program.cs (or a bootstrap extension):
-//   builder.Services.AddMemoryCache();
-//   builder.Services.AddScoped<IPromptRepository, PromptRepository>();
-//
-// The DbContext must expose:  public DbSet<AgentPrompt> AgentPrompts { get; set; }
-// Replace YourDbContext below with the actual DbContext type of the project.
-
-/// <summary>
-/// Reads the active agent prompt from the database with an in-memory cache
-/// (default TTL: 5 minutes sliding). Cache invalidation is time-based;
-/// for immediate invalidation after a prompt update call
-/// <c>IMemoryCache.Remove($"agent_prompt:{agentKey}")</c>.
-/// </summary>
-public sealed class PromptRepository(YourDbContext db, IMemoryCache cache) : IPromptRepository
-{
-    private static readonly TimeSpan CacheTtl = TimeSpan.FromMinutes(5);
-
-    public async Task<string> GetActivePromptAsync(string agentKey, CancellationToken ct = default)
-    {
-        var cacheKey = $"agent_prompt:{agentKey}";
-
-        if (cache.TryGetValue(cacheKey, out string? cached) && cached is not null)
-            return cached;
-
-        var content = await db.AgentPrompts
-            .Where(p => p.AgentKey == agentKey && p.IsActive)
-            .Select(p => p.Content)
-            .FirstOrDefaultAsync(ct);
-
-        if (content is null)
-            throw new InvalidOperationException(
-                $"No active prompt found for agent '{agentKey}'. " +
-                "Insert a row in agent_prompts with is_active = true for this agent_key.");
-
-        cache.Set(cacheKey, content, new MemoryCacheEntryOptions
-        {
-            SlidingExpiration = CacheTtl
-        });
-
-        return content;
-    }
-}
+// TEMPLATE: PromptRepository.cs
+// PURPOSE: Loads agent system prompts from the agent_prompts table, with in-memory caching.
+// READS STANDARD: ai-agents-prompt-sources
+// PLACEHOLDERS:
+//   {{PROJECT_NAMESPACE}}  — root namespace of the project
+// LAST-VERIFIED: 2026-05-19
+
+using Microsoft.EntityFrameworkCore;
+using Microsoft.Extensions.Caching.Memory;
+
+namespace {{PROJECT_NAMESPACE}}.AI;
+
+// ---------------------------------------------------------------------------
+// Entity — map this to the agent_prompts table (EF Core convention mapping).
+// Alternatively, configure via Fluent API in your DbContext.OnModelCreating.
+// ---------------------------------------------------------------------------
+
+/// <summary>
+/// EF Core entity for the agent_prompts table.
+/// Content is immutable once written — use INSERT + flip is_active for versioning.
+/// </summary>
+public sealed class AgentPrompt
+{
+    public Guid Id { get; init; } = Guid.CreateVersion7();
+    public required string AgentKey { get; init; }
+    public int Version { get; init; }
+    public required string Content { get; init; }
+    public bool IsActive { get; set; }
+    public DateTimeOffset CreatedAt { get; init; } = DateTimeOffset.UtcNow;
+}
+
+// ---------------------------------------------------------------------------
+// Interface
+// ---------------------------------------------------------------------------
+
+/// <summary>
+/// Provides the active system-prompt content for a named agent.
+/// </summary>
+public interface IPromptRepository
+{
+    /// <summary>
+    /// Returns the content of the currently active prompt for <paramref name="agentKey"/>.
+    /// </summary>
+    /// <exception cref="InvalidOperationException">
+    /// Thrown when no row with <c>is_active = true</c> exists for <paramref name="agentKey"/>.
+    /// </exception>
+    Task<string> GetActivePromptAsync(string agentKey, CancellationToken ct = default);
+}
+
+// ---------------------------------------------------------------------------
+// Implementation
+// ---------------------------------------------------------------------------
+
+// NOTE: Register in Program.cs (or a bootstrap extension):
+//   builder.Services.AddMemoryCache();
+//   builder.Services.AddScoped<IPromptRepository, PromptRepository>();
+//
+// The DbContext must expose:  public DbSet<AgentPrompt> AgentPrompts { get; set; }
+// Replace YourDbContext below with the actual DbContext type of the project.
+
+/// <summary>
+/// Reads the active agent prompt from the database with an in-memory cache
+/// (default TTL: 5 minutes sliding). Cache invalidation is time-based;
+/// for immediate invalidation after a prompt update call
+/// <c>IMemoryCache.Remove($"agent_prompt:{agentKey}")</c>.
+/// </summary>
+public sealed class PromptRepository(YourDbContext db, IMemoryCache cache) : IPromptRepository
+{
+    private static readonly TimeSpan CacheTtl = TimeSpan.FromMinutes(5);
+
+    public async Task<string> GetActivePromptAsync(string agentKey, CancellationToken ct = default)
+    {
+        var cacheKey = $"agent_prompt:{agentKey}";
+
+        if (cache.TryGetValue(cacheKey, out string? cached) && cached is not null)
+            return cached;
+
+        var content = await db.AgentPrompts
+            .Where(p => p.AgentKey == agentKey && p.IsActive)
+            .Select(p => p.Content)
+            .FirstOrDefaultAsync(ct);
+
+        if (content is null)
+            throw new InvalidOperationException(
+                $"No active prompt found for agent '{agentKey}'. " +
+                "Insert a row in agent_prompts with is_active = true for this agent_key.");
+
+        cache.Set(cacheKey, content, new MemoryCacheEntryOptions
+        {
+            SlidingExpiration = CacheTtl
+        });
+
+        return content;
+    }
+}

package/framework/templates/code/sql/ai-agents/agent-prompts-migration.sql.template

@@ -1,58 +1,58 @@
--- TEMPLATE: agent-prompts-migration.sql
--- PURPOSE: Creates the agent_prompts table for runtime-editable, versioned agent instructions.
--- READS STANDARD: ai-agents-prompts-in-db
--- PLACEHOLDERS:
---   {{SCHEMA}}  — target schema (default: public)
--- LAST-VERIFIED: 2026-05-19
-
--- ============================================================
--- 1. TABLE
--- ============================================================
-
-CREATE TABLE {{SCHEMA}}.agent_prompts (
-    id          uuid        PRIMARY KEY DEFAULT gen_random_uuid(),
-    agent_key   text        NOT NULL,
-    version     int         NOT NULL,
-    content     text        NOT NULL,
-    is_active   boolean     NOT NULL DEFAULT false,
-    created_at  timestamptz NOT NULL DEFAULT now()
-);
-
-COMMENT ON TABLE  {{SCHEMA}}.agent_prompts              IS 'Runtime-editable, versioned system prompts for AI agents.';
-COMMENT ON COLUMN {{SCHEMA}}.agent_prompts.agent_key    IS 'Logical identifier for the agent (e.g. ''proposal-agent'').';
-COMMENT ON COLUMN {{SCHEMA}}.agent_prompts.version      IS 'Monotonically increasing per agent_key. Never reuse a version number.';
-COMMENT ON COLUMN {{SCHEMA}}.agent_prompts.content      IS 'Immutable system-prompt text. UPDATE of this column is forbidden — insert a new row instead.';
-COMMENT ON COLUMN {{SCHEMA}}.agent_prompts.is_active    IS 'Exactly one row per agent_key may have is_active = true (enforced by ix_agent_prompts_active).';
-COMMENT ON COLUMN {{SCHEMA}}.agent_prompts.created_at   IS 'Audit timestamp. Set at insert time, never updated.';
-
--- ============================================================
--- 2. INDEXES
--- ============================================================
-
--- Unique partial index: enforces at DB level that only one active prompt
--- exists per agent_key. Any attempt to INSERT a second active row for the
--- same key will fail with a unique-constraint violation.
-CREATE UNIQUE INDEX ix_agent_prompts_active
-    ON {{SCHEMA}}.agent_prompts (agent_key)
-    WHERE is_active;
-
--- Regular index for the PromptRepository lookup (agent_key + is_active filter).
-CREATE INDEX ix_agent_prompts_agent_key
-    ON {{SCHEMA}}.agent_prompts (agent_key);
-
--- Unique index: prevents duplicate version numbers per agent_key.
-CREATE UNIQUE INDEX ix_agent_prompts_agent_key_version
-    ON {{SCHEMA}}.agent_prompts (agent_key, version);
-
--- ============================================================
--- 3. SEED (optional — remove if managed via backoffice)
--- ============================================================
-
--- INSERT INTO {{SCHEMA}}.agent_prompts (agent_key, version, content, is_active)
--- VALUES ('my-agent', 1, 'You are a helpful assistant.', true);
-
--- ============================================================
--- ROLLBACK
--- ============================================================
-
--- DROP TABLE IF EXISTS {{SCHEMA}}.agent_prompts CASCADE;
+-- TEMPLATE: agent-prompts-migration.sql
+-- PURPOSE: Creates the agent_prompts table for runtime-editable, versioned agent instructions.
+-- READS STANDARD: ai-agents-prompt-sources
+-- PLACEHOLDERS:
+--   {{SCHEMA}}  — target schema (default: public)
+-- LAST-VERIFIED: 2026-05-19
+
+-- ============================================================
+-- 1. TABLE
+-- ============================================================
+
+CREATE TABLE {{SCHEMA}}.agent_prompts (
+    id          uuid        PRIMARY KEY DEFAULT gen_random_uuid(),
+    agent_key   text        NOT NULL,
+    version     int         NOT NULL,
+    content     text        NOT NULL,
+    is_active   boolean     NOT NULL DEFAULT false,
+    created_at  timestamptz NOT NULL DEFAULT now()
+);
+
+COMMENT ON TABLE  {{SCHEMA}}.agent_prompts              IS 'Runtime-editable, versioned system prompts for AI agents.';
+COMMENT ON COLUMN {{SCHEMA}}.agent_prompts.agent_key    IS 'Logical identifier for the agent (e.g. ''proposal-agent'').';
+COMMENT ON COLUMN {{SCHEMA}}.agent_prompts.version      IS 'Monotonically increasing per agent_key. Never reuse a version number.';
+COMMENT ON COLUMN {{SCHEMA}}.agent_prompts.content      IS 'Immutable system-prompt text. UPDATE of this column is forbidden — insert a new row instead.';
+COMMENT ON COLUMN {{SCHEMA}}.agent_prompts.is_active    IS 'Exactly one row per agent_key may have is_active = true (enforced by ix_agent_prompts_active).';
+COMMENT ON COLUMN {{SCHEMA}}.agent_prompts.created_at   IS 'Audit timestamp. Set at insert time, never updated.';
+
+-- ============================================================
+-- 2. INDEXES
+-- ============================================================
+
+-- Unique partial index: enforces at DB level that only one active prompt
+-- exists per agent_key. Any attempt to INSERT a second active row for the
+-- same key will fail with a unique-constraint violation.
+CREATE UNIQUE INDEX ix_agent_prompts_active
+    ON {{SCHEMA}}.agent_prompts (agent_key)
+    WHERE is_active;
+
+-- Regular index for the PromptRepository lookup (agent_key + is_active filter).
+CREATE INDEX ix_agent_prompts_agent_key
+    ON {{SCHEMA}}.agent_prompts (agent_key);
+
+-- Unique index: prevents duplicate version numbers per agent_key.
+CREATE UNIQUE INDEX ix_agent_prompts_agent_key_version
+    ON {{SCHEMA}}.agent_prompts (agent_key, version);
+
+-- ============================================================
+-- 3. SEED (optional — remove if managed via backoffice)
+-- ============================================================
+
+-- INSERT INTO {{SCHEMA}}.agent_prompts (agent_key, version, content, is_active)
+-- VALUES ('my-agent', 1, 'You are a helpful assistant.', true);
+
+-- ============================================================
+-- ROLLBACK
+-- ============================================================
+
+-- DROP TABLE IF EXISTS {{SCHEMA}}.agent_prompts CASCADE;

package/framework/templates/dotnet/ai-kit/Directory.Build.props

@@ -0,0 +1,25 @@
+<Project>
+
+  <!--
+    Propriedades comuns a TODOS os projetos do Morph.AiKit.
+
+    NAO escreva versao de pacote aqui nem em nenhum .csproj: as versoes vivem em
+    Directory.Packages.props, GERADO por scripts/generate-ai-pin-props.js a partir de
+    framework/ai-pin.json. Com Central Package Management ligado, um <PackageReference>
+    com atributo Version vira erro de restore (NU1008) — e por isso que um .csproj nao
+    consegue divergir do pin.
+  -->
+  <PropertyGroup>
+    <TargetFramework>net10.0</TargetFramework>
+    <LangVersion>latest</LangVersion>
+    <Nullable>enable</Nullable>
+    <ImplicitUsings>enable</ImplicitUsings>
+
+    <!-- Um kit cujo proposito e gritar nao pode engolir warning. -->
+    <TreatWarningsAsErrors>true</TreatWarningsAsErrors>
+
+    <Deterministic>true</Deterministic>
+    <IsPackable>false</IsPackable>
+  </PropertyGroup>
+
+</Project>

package/framework/templates/dotnet/ai-kit/Directory.Packages.props

@@ -0,0 +1,26 @@
+<Project>
+  <!-- GERADO por scripts/generate-ai-pin-props.js a partir de framework/ai-pin.json. Não editar à mão — mexer no pin é mexer em framework/ai-pin.json. -->
+  <!-- pinnedAt: 2026-09-08 · fonte: https://api.nuget.org/v3/registration5-semver1/{id}/index.json -->
+  <PropertyGroup>
+    <ManagePackageVersionsCentrally>true</ManagePackageVersionsCentrally>
+  </PropertyGroup>
+
+  <!-- Pacotes de IA (framework/ai-pin.json → packages) -->
+  <ItemGroup>
+    <PackageVersion Include="Microsoft.Agents.AI" Version="1.20.0" />
+    <PackageVersion Include="Microsoft.Extensions.AI" Version="10.9.0" />
+    <PackageVersion Include="OpenAI" Version="2.13.0" />
+    <PackageVersion Include="Google.GenAI" Version="1.21.0" />
+    <PackageVersion Include="ModelContextProtocol" Version="2.2.0" />
+    <PackageVersion Include="Microsoft.Extensions.AI.Evaluation" Version="10.9.0" />
+    <PackageVersion Include="Microsoft.Extensions.AI.Evaluation.Quality" Version="10.9.0" />
+    <PackageVersion Include="Microsoft.Extensions.AI.Evaluation.Reporting" Version="10.9.0" />
+  </ItemGroup>
+
+  <!-- Build e teste (framework/ai-pin.json → toolchain) -->
+  <ItemGroup>
+    <PackageVersion Include="Microsoft.NET.Test.Sdk" Version="18.9.0" />
+    <PackageVersion Include="xunit.v3" Version="4.0.0" />
+    <PackageVersion Include="xunit.runner.visualstudio" Version="4.0.0" />
+  </ItemGroup>
+</Project>

package/framework/templates/dotnet/ai-kit/Morph.AiKit.sln

@@ -0,0 +1,56 @@
+
+Microsoft Visual Studio Solution File, Format Version 12.00
+# Visual Studio Version 17
+VisualStudioVersion = 17.0.31903.59
+MinimumVisualStudioVersion = 10.0.40219.1
+Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "src", "src", "{827E0CD3-B72D-47B6-A68D-7590B98EB39B}"
+EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "Morph.AiKit", "src\Morph.AiKit\Morph.AiKit.csproj", "{0577D911-1F1E-422E-A1CD-23379F308B05}"
+EndProject
+Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "tests", "tests", "{0AB3BF05-4346-4AA6-1389-037BE0695223}"
+EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "Morph.AiKit.Tests", "tests\Morph.AiKit.Tests\Morph.AiKit.Tests.csproj", "{A0543CA3-CC90-45F6-940E-3F58570BB4D3}"
+EndProject
+Global
+	GlobalSection(SolutionConfigurationPlatforms) = preSolution
+		Debug|Any CPU = Debug|Any CPU
+		Debug|x64 = Debug|x64
+		Debug|x86 = Debug|x86
+		Release|Any CPU = Release|Any CPU
+		Release|x64 = Release|x64
+		Release|x86 = Release|x86
+	EndGlobalSection
+	GlobalSection(ProjectConfigurationPlatforms) = postSolution
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Debug|x64.ActiveCfg = Debug|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Debug|x64.Build.0 = Debug|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Debug|x86.ActiveCfg = Debug|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Debug|x86.Build.0 = Debug|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Release|Any CPU.Build.0 = Release|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Release|x64.ActiveCfg = Release|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Release|x64.Build.0 = Release|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Release|x86.ActiveCfg = Release|Any CPU
+		{0577D911-1F1E-422E-A1CD-23379F308B05}.Release|x86.Build.0 = Release|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Debug|x64.ActiveCfg = Debug|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Debug|x64.Build.0 = Debug|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Debug|x86.ActiveCfg = Debug|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Debug|x86.Build.0 = Debug|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Release|Any CPU.Build.0 = Release|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Release|x64.ActiveCfg = Release|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Release|x64.Build.0 = Release|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Release|x86.ActiveCfg = Release|Any CPU
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3}.Release|x86.Build.0 = Release|Any CPU
+	EndGlobalSection
+	GlobalSection(SolutionProperties) = preSolution
+		HideSolutionNode = FALSE
+	EndGlobalSection
+	GlobalSection(NestedProjects) = preSolution
+		{0577D911-1F1E-422E-A1CD-23379F308B05} = {827E0CD3-B72D-47B6-A68D-7590B98EB39B}
+		{A0543CA3-CC90-45F6-940E-3F58570BB4D3} = {0AB3BF05-4346-4AA6-1389-037BE0695223}
+	EndGlobalSection
+EndGlobal