feat(workflow-executor): finalize scaffold — clean CLAUDE.md, remove premature deps, add smoke test

- Rewrite CLAUDE.md with project overview and architecture principles, remove changelog
- Remove unused dependencies (ai-proxy, sequelize, zod) per YAGNI
- Add smoke test so CI passes

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: alban bertolini

packages/workflow-executor/CLAUDE.md

@@ -1,21 +1,25 @@
 # @forestadmin/workflow-executor
 
-Bibliothèque TypeScript framework-agnostic qui exécute des steps de workflow côté client (infra du client, à côté de l'agent Forest Admin).
+> **Note to Claude**: Keep this file up to date. When adding a new feature, module, architectural pattern, or dependency, update the relevant section below.
 
-## Architecture
+## Overview
 
-- **Pull-based** — L'executor poll `WorkflowPort`. `triggerPoll(runId)` pour accélérer un run spécifique
-- **Atomic** — Chaque step exécutée en isolation. `RunStore` assure la continuité entre steps
-- **Privacy** — Zéro donnée client dans l'orchestrateur. Données dans `RunStore`
-- **Ports** — Toute IO passe par une interface injectée
-- **AI intégré** — Utilise `@forestadmin/ai-proxy` (Router) pour créer les modèles et charger les remote tools
+TypeScript library (framework-agnostic) that executes workflow steps on the client's infrastructure, alongside the Forest Admin agent. The orchestrator never sees client data — it only sends step definitions; this package fetches them and runs them locally.
+
+## Architecture Principles
+
+- **Pull-based** — The executor polls for pending steps via `WorkflowPort`. `triggerPoll(runId)` fast-tracks a specific run.
+- **Atomic** — Each step is executed in isolation. `RunStore` maintains continuity between steps.
+- **Privacy** — Zero client data leaves the client's infrastructure. All data lives in `RunStore`.
+- **Ports (IO injection)** — Every external IO goes through an injected port interface, making the core pure and testable.
+- **AI integration** — Uses `@forestadmin/ai-proxy` (Router) to create models and load remote tools.
 
 ## Commands
 
 ```bash
 yarn workspace @forestadmin/workflow-executor build      # Build
-yarn workspace @forestadmin/workflow-executor test        # Run tests
-yarn workspace @forestadmin/workflow-executor lint        # Lint
+yarn workspace @forestadmin/workflow-executor test       # Run tests
+yarn workspace @forestadmin/workflow-executor lint       # Lint
 ```
 
 ## Testing
@@ -24,11 +28,3 @@ yarn workspace @forestadmin/workflow-executor lint        # Lint
 - Use AAA pattern (Arrange, Act, Assert)
 - Test behavior, not implementation
 - Strong assertions: verify exact arguments, not just that a function was called
-
-## Changelog
-
-> **IMPORTANT**: When a new feature, fix, or change is implemented, add an entry below summarizing what was done.
-
-| Date       | Type    | Summary |
-|------------|---------|---------|
-| 2026-03-16 | setup   | Initial package scaffolding (package.json, tsconfig, jest, CI integration) |

packages/workflow-executor/package.json

@@ -11,11 +11,6 @@
     "url": "git+https://github.com/ForestAdmin/agent-nodejs.git",
     "directory": "packages/workflow-executor"
   },
-  "dependencies": {
-    "@forestadmin/ai-proxy": "1.5.0",
-    "sequelize": "^6.37.5",
-    "zod": "^4.3.5"
-  },
   "files": [
     "dist/**/*.js",
     "dist/**/*.d.ts"

packages/workflow-executor/test/index.test.ts

@@ -0,0 +1,5 @@
+describe('workflow-executor', () => {
+  it('should be importable', () => {
+    expect(require('../src/index')).toBeDefined();
+  });
+});