feat(domain): add messaging remote control service by arielshad · Pull Request #512 · shep-ai/shep

arielshad · 2026-04-02T09:20:03Z

Summary

Add TypeSpec domain models for messaging concepts (commands, config, notifications, platform enums)
Implement messaging service infrastructure with tunnel adapter, chat relay, command executor, content sanitizer, and notification emitter
Add CLI commands for messaging configuration (enable/disable, platform config, notification settings)
Register messaging service in DI container and integrate with _serve command daemon lifecycle

What Changed

TypeSpec Models (tsp/)

New enums: MessagingCommandType, MessagingFrameType, MessagingPlatform
New entities: MessagingCommand, MessagingConfig, MessagingPlatformConfig, MessagingNotification
Extended Settings model with messaging configuration

Messaging Service (packages/core/src/infrastructure/services/messaging/)

messaging.service.ts — Orchestrates tunnel connection, chat relay, command execution, and notifications
messaging-tunnel.adapter.ts — WebSocket tunnel adapter for commands-com gateway connectivity
chat-relay.ts — Relays chat messages between external platforms and shep sessions
command-executor.ts — Executes remote commands received via messaging with permission controls
content-sanitizer.ts — Sanitizes message content for safe handling
notification-emitter.ts — Emits notifications to connected messaging platforms

CLI Commands (src/presentation/cli/commands/settings/messaging.command.ts)

settings messaging enable/disable — Toggle messaging remote control
settings messaging platform — Configure platform-specific settings
settings messaging notifications — Configure notification preferences

Integration

DI container registration for messaging service and sub-components
_serve command integration for daemon lifecycle management

Evidence

Evidence	Source
39 messaging unit tests passing (chat-relay, command-executor, content-sanitizer, notification-emitter)	`specs/082-messaging-remote-control/evidence/messaging-unit-tests.txt`
12 serve command tests passing with messaging integration	`specs/082-messaging-remote-control/evidence/serve-command-tests.txt`
Full test suite — 378 files, 5338 tests, zero failures	`specs/082-messaging-remote-control/evidence/full-unit-test-summary.txt`
TypeScript build completes successfully	`specs/082-messaging-remote-control/evidence/build-output.txt`
TypeSpec compilation succeeds for all messaging models	`specs/082-messaging-remote-control/evidence/tsp-compile-output.txt`

Test Plan

All 39 messaging service unit tests pass
All 12 serve command tests pass (including messaging lifecycle)
Full unit test suite passes (5338 tests, 0 failures)
Build compiles without errors
Lint passes with zero warnings

Built with Shep 🐑 Shep Bot

github-actions · 2026-04-02T09:20:35Z

🔒 Security Issues Detected

Scanner	Status	Details
Gitleaks	❌ Failed	Secret detection in git history
Semgrep	✅ Pass	SAST for TypeScript/JavaScript

⚠️ Please review the failed checks above before merging.

_{🤖 Generated by CI Security Scan | View Details}

github-actions · 2026-04-02T09:28:15Z

🔒 Security Issues Detected

Scanner	Status	Details
Gitleaks	❌ Failed	Secret detection in git history
Semgrep	✅ Pass	SAST for TypeScript/JavaScript

⚠️ Please review the failed checks above before merging.

_{🤖 Generated by CI Security Scan | View Details}

github-actions · 2026-04-02T09:42:27Z

Dev Release Published

Artifact	Version	Install
npm	`1.163.0-pr512.9a94785`	`npm install -g @shepai/cli@1.163.0-pr512.9a94785`

_{Published from commit 5ae742a | View CI}

## Summary After migrating to the Anthropic V2 SDK for interactive chat sessions, the `permissionMode: 'bypassPermissions'` option stopped working because the V2 API hardcodes `allowDangerouslySkipPermissions: false`. This caused the agent to silently fall back to default permission mode, making it unable to execute Bash commands and other tools without user approval prompts. **Root cause:** V2 SDK ignores `bypassPermissions` permission mode, so all tool calls were being blocked or prompted. **Fix:** Replace `permissionMode: 'bypassPermissions'` with the V2-supported `allowedTools` array, which pre-approves all standard Claude Code tools (Bash, Read, Write, Edit, Glob, Grep, Agent, etc.) at the CLI level. `AskUserQuestion` is intentionally excluded from the auto-allow list so it continues to be intercepted by the `canUseTool` callback for user interaction. ## Changes - Added `AUTO_ALLOWED_TOOLS` constant listing all 26 standard Claude Code tools to auto-allow - Replaced conditional `permissionMode: 'bypassPermissions'` with unconditional `allowedTools` pass-through - `canUseTool` callback is now only passed when `onUserQuestion` is provided (no fallback to bypassPermissions) - Added comprehensive unit tests (11 test cases) covering: - `allowedTools` is passed with all standard tools - `AskUserQuestion` is excluded from auto-allowed tools - `permissionMode` is no longer set - `canUseTool` callback behavior for both regular tools and `AskUserQuestion` - Environment variable stripping, system prompt forwarding, default model ## Test plan - [x] Unit tests pass (11/11) for the interactive executor - [ ] Manual verification: run `shep` interactive chat and confirm Bash/tool execution works without permission prompts - [ ] Verify `AskUserQuestion` still pauses for user input Built with Shep 🐑 [Shep Bot](https://github.com/shep-ai/shep) --------- Co-authored-by: shep-ai[bot] <bot@shep.bot> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

## [1.175.1](v1.175.0...v1.175.1) (2026-04-06) ### Bug Fixes * **agents:** use allowedtools for v2 sdk tool permissions ([#534](#534)) ([7c9c50e](7c9c50e))

Add TypeSpec models for external messaging integration via the Commands.com Gateway: MessagingPlatform enum, MessagingCommand, MessagingNotification, and MessagingConfig settings model. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Implement external messaging remote control via Commands.com Gateway, enabling Telegram/WhatsApp-based notifications, commands, and chat relay for managing Shep features remotely. Changes: - IMessagingService port interface for clean architecture boundary - MessagingTunnelAdapter: WebSocket tunnel to Gateway with heartbeat and auto-reconnect - MessagingCommandExecutor: maps /new, /approve, /reject, /stop, /resume, /status commands to existing use cases - MessagingNotificationEmitter: subscribes to NotificationEventBus, debounces and sanitizes events before forwarding - MessagingChatRelay: bidirectional agent session relay with output buffering to avoid message flooding - ContentSanitizer: strips paths, env vars, code blocks from outbound messages to prevent leaking sensitive content - MessagingService orchestrator: coordinates all components - DI container registration with lazy proxy for zero-cost CLI startup - CLI: shep settings messaging wizard (connect/status/disconnect) - Daemon: start/stop messaging service in _serve command - 39 unit tests covering all components Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

…x spec yaml Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

… scan The content-sanitizer test used 'sk-abc123def456' as a dummy API key value, which gitleaks flagged as a generic-api-key secret. Replaced with 'test-value' which still exercises the env-var stripping logic without triggering the secret scanner.

Replace the hand-rolled messaging scaffolding with a working end-to-end integration against the Commands.com Gateway. Users can now pair a Telegram bot via the web UI or CLI, point their bot webhook at the gateway's public ingress URL, and drive Shep from their phone with slash commands plus an interactive /chat relay that streams agent output back over Telegram. Core additions: - IGatewayClient + HttpGatewayClient for OAuth token fetch and integration route registration - Tunnel adapter rewritten with the ws library so bearer headers work on upgrade, speaking the real tunnel.activate/request/response protocol - IMessageSender output port decouples outbound delivery from inbound tunnel frames; TelegramMessageSender implements it via HttpTelegramClient - Webhook parsers for Telegram Update and WhatsApp Business Cloud payloads - Begin/Confirm/Disconnect pairing use cases shared by CLI and Web - Interactive /chat relay subscribes to IInteractiveSessionService and forwards deltas back to chat; free-form text during a relay goes to sendUserMessage - Per-platform botToken in settings with env var fallback - Web settings section, storybook stories, e2e test, and local setup docs Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Refresh the allowlist entry for the fake api key in content-sanitizer test to reference the rebased commit sha so gitleaks scans pass. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Feature 082 shipped with several silent gaps that prevented the Telegram remote control flow from ever completing a real pairing. This change closes all of them so /pair, /list, /help actually round-trip between the user's chat and the Shep daemon. - add migration 056 with 24 nullable messaging_* settings columns; safe defaults mean older builds keep working untouched - extend sqlite settings mapper + repository to round-trip messagingconfig - relax isconfigured so the tunnel starts in pending-pairing state and can auto-confirm /pair <code> over the wire - fix tunnel.activate frame shape to send a batched routes array and parse the gateway's results array response (prior code sent the wrong singular shape so no routes were ever activated) - gate messaging startup in dev-server.ts behind shep-enable-messaging - add 4 round-trip persistence tests and update tunnel adapter tests; 96 targeted tests pass Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Resolve merge conflicts from main (v1.181.0) including: - DI container: keep both messaging service and application/terminal imports - TypeSpec: include both messaging.tsp and application.tsp entity imports - Generated output.ts: preserve messaging types alongside application types - Interactive executor test: adopt main's V2 SDK systemPrompt behavior - Add missing messaging translation key for Ukrainian locale Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

BeginMessagingPairingUseCase calls HttpGatewayClient which makes real HTTP requests to a Commands.com Gateway that does not exist in the E2E environment. The pairing server action silently fails, so the dialog never opens and the test times out waiting for messaging-pairing-dialog. Add StubGatewayClient that returns deterministic fake responses and register it via SHEP_MOCK_GATEWAY=1 env var (same pattern as SHEP_MOCK_EXECUTOR). Pass the env var in playwright.config.ts so the dev server uses the stub during E2E runs. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

github-actions · 2026-04-12T19:26:54Z

Dev Release Published

Artifact	Version	Install
npm	`1.181.0-pr512.ba01d33`	`npm install -g @shepai/cli@1.181.0-pr512.ba01d33`

_{Published from commit 2529d68 | View CI}

github-actions · 2026-04-13T08:03:26Z

Dev Release Published

Artifact	Version	Install
npm	`1.182.0-pr512.369be8e`	`npm install -g @shepai/cli@1.182.0-pr512.369be8e`

_{Published from commit 3ae33e6 | View CI}

Co-Authored-By: Shep Bot <shep-agent@users.noreply.github.com>

github-actions · 2026-04-26T17:23:09Z

Dev Release Published

Artifact	Version	Install
npm	`1.193.0-pr512.745ede6`	`npm install -g @shepai/cli@1.193.0-pr512.745ede6`

_{Published from commit e37c020 | View CI}

blackpc and others added 12 commits April 6, 2026 16:47

chore(release): 1.175.1 [skip ci]

a8edb49

## [1.175.1](v1.175.0...v1.175.1) (2026-04-06) ### Bug Fixes * **agents:** use allowedtools for v2 sdk tool permissions ([#534](#534)) ([7c9c50e](7c9c50e))

fix(domain): add messaging service mock to serve command tests and fi…

ffdb594

…x spec yaml Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

chore(agents): capture evidence for messaging remote control feature

90318bf

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

chore(agents): capture evidence for messaging remote control feature

f3682c1

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

chore(agents): capture evidence for messaging remote control feature

cea80fd

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

chore(specs): update feature yaml with completed phases tracking

42c9b9d

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

fix(ci): update gitleaksignore sha after rebase on main

d333ae0

Refresh the allowlist entry for the fake api key in content-sanitizer test to reference the rebased commit sha so gitleaks scans pass. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

arielshad force-pushed the feat/messaging-remote-control branch 2 times, most recently from 42b1f72 to d333ae0 Compare April 6, 2026 15:48

arielshad and others added 3 commits April 9, 2026 19:12

fix(domain): merge main into feat/messaging-remote-control

3ae33e6

arielshad and others added 2 commits April 26, 2026 19:56

fix(domain): merge main into feat/messaging-remote-control

6e5deaf

Co-Authored-By: Shep Bot <shep-agent@users.noreply.github.com>

fix(ci): attempt 1/10 — add ws to electron package dependencies

e37c020

Co-Authored-By: Shep Bot <shep-agent@users.noreply.github.com>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(domain): add messaging remote control service#512

feat(domain): add messaging remote control service#512
arielshad wants to merge 18 commits intomainfrom
feat/messaging-remote-control

arielshad commented Apr 2, 2026

Uh oh!

github-actions Bot commented Apr 2, 2026

Uh oh!

github-actions Bot commented Apr 2, 2026

Uh oh!

github-actions Bot commented Apr 2, 2026

Uh oh!

github-actions Bot commented Apr 12, 2026

Uh oh!

github-actions Bot commented Apr 13, 2026

Uh oh!

github-actions Bot commented Apr 26, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

arielshad commented Apr 2, 2026

Summary

What Changed

Evidence

Test Plan

Uh oh!

github-actions Bot commented Apr 2, 2026

🔒 Security Issues Detected

Uh oh!

github-actions Bot commented Apr 2, 2026

🔒 Security Issues Detected

Uh oh!

github-actions Bot commented Apr 2, 2026

Dev Release Published

Uh oh!

github-actions Bot commented Apr 12, 2026

Dev Release Published

Uh oh!

github-actions Bot commented Apr 13, 2026

Dev Release Published

Uh oh!

github-actions Bot commented Apr 26, 2026

Dev Release Published

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants