feat: Feature parity for remote mode

[kdy1]

Implement remote mode support for previously unimplemented Tauri commands.

Summary

• Add CancelTask, GetCompositeTaskNodes, GetTaskLogs RPC types
• Add 5 new server endpoints: cancel, dismiss-approval, create-pr, get-composite-nodes, get-logs
• Add corresponding remote client methods
• Update Tauri cancel_task to call remote API instead of returning error
• Update Tauri get_composite_task_nodes to fetch from server
• Update Tauri get_task_logs to stream logs from server
• Update API documentation

Closes #252

Generated with Claude Code

[kdy1]

@claude Fix all critical issues in https://github.com/delinoio/delidev/pull/253#issuecomment-3866230758

[kdy1]

@claude fix all critical issues in https://github.com/delinoio/delidev/pull/253#issuecomment-3866358070

[kdy1]

@claude Fix critical issues in https://github.com/delinoio/delidev/pull/253#issuecomment-3869607422

[kdy1]

@claude Fix all critical issues in https://github.com/delinoio/delidev/pull/253#issuecomment-3869660787

[claude]

PR Review: Feature Parity for Remote Mode

I've reviewed PR #253 and overall this is a well-implemented feature that brings important remote mode capabilities. Here are my findings:

✅ Strengths

1. Excellent Security Implementation

   • Strong SSRF protection in validate_worker_endpoint_url() (worker_registry.rs:18-67)
   • Proper validation at registration time AND before making requests (defense-in-depth)
   • Comprehensive test coverage for SSRF validation
   • Good error messages that don't leak internal details

2. Well-Structured Code

   • Clear separation of concerns between server endpoints and client methods
   • Consistent error handling patterns
   • Good use of constants for limits (MAX_AFTER_EVENT_ID, MAX_LOG_LINES_PER_SESSION)
   • Extensive test coverage with 20+ new tests

3. Documentation

   • API documentation updated in docs/api.md
   • Design document updated to reflect cancellation workflow
   • Clear inline comments explaining non-obvious logic

🔧 Issues to Address

1. CRITICAL: IPv6 Private Network Detection Incomplete (worker_registry.rs:80-84)

The IPv6 validation is missing critical ranges that could enable SSRF:

• fc00::/7 - Unique Local Addresses (IPv6 equivalent of 10.0.0.0/8)
• fe80::/10 - Link-local addresses
• IPv4-mapped IPv6 addresses like ::ffff:192.168.0.1

2. Race Condition in Task Cancellation (api/task.rs:108-189)

The code checks task status before signaling the worker, but the task could transition between the check and the update. If the worker completes the task between the check and the update, you'll mark a completed task as cancelled.

Recommendation: Use a database transaction or compare-and-swap operation to ensure atomicity.

3. Potential Information Leak (api/task.rs:128-140)

Logging the full endpoint URL could expose internal infrastructure details in logs. Sanitize the URL in logs or log only the validation error.

4. Unbounded Resource Usage (api/task.rs:315-327)

N+1 query pattern in get_composite_task_nodes. For a composite task with 1000 nodes, this makes 1000 database queries. Consider adding a batch fetch method to TaskStore.

5. Client-Side Log Parsing Could Fail Silently (commands/task.rs:1024-1032)

Parse failures are only logged at debug level. Log at warn level at minimum or consider adding a metric.

⚠️ Concerns

1. HTTP Client Pool Configuration - HTTP_POOL_MAX_IDLE_PER_HOST: 5 seems low for many workers
2. No Retry Logic for Worker Communication - A single network glitch causes cancellation failure
3. Test Coverage Gaps - No tests for concurrent cancellation or timeout behavior
4. Error Message Quality - Using {:?} in user-facing errors exposes internal enum representation

🎯 Recommendation

Approve with changes required for the IPv6 SSRF vulnerability. The other issues are important but not blocking.

1. MUST FIX before merge: IPv6 private network detection
2. SHOULD FIX before merge: Race condition in cancellation, N+1 queries
3. NICE TO HAVE: Other suggestions can be follow-up PRs

Great work overall! The security-first approach and comprehensive testing are particularly commendable.