Skip to content

Commit 1a34ab2

Browse files
committed
chore: remove TODO johannesjo#13 and johannesjo#14 sections (already landed)
1 parent e8673ae commit 1a34ab2

1 file changed

Lines changed: 0 additions & 16 deletions

File tree

TODOS.md

Lines changed: 0 additions & 16 deletions
Original file line numberDiff line numberDiff line change
@@ -9,22 +9,6 @@
99

1010
---
1111

12-
## Beta blockers — lifecycle/scoping issues
13-
14-
### 13. Coordinator cross-contamination — tasks not scoped to their coordinator
15-
16-
**Files:** `electron/mcp/server.ts:75-80` (`create_task`), `electron/remote/server.ts:300-325,420-450` (`list_tasks`, `send_prompt`, `merge_task`, `close_task`)
17-
**What's wrong:** `create_task` sends no `projectId`. Tool handlers accept `taskId` without verifying the caller's `coordinatorId`. Multiple concurrent coordinators can see and control each other's sub-tasks.
18-
**Done when:** Every MCP tool call that targets a specific task passes `coordinatorId`, and each handler rejects calls where `task.coordinatorTaskId !== coordinatorId`.
19-
20-
### 14. Coordinator MCP broken when remote access server was already running
21-
22-
**File:** `electron/ipc/register.ts` (`StartMCPServer`), `electron/remote/server.ts` (`startRemoteServer`)
23-
**What's wrong:** `StartMCPServer` skips `startRemoteServer()` if already running. If remote access started first (without a coordinator), coordinator-specific API routes are never registered.
24-
**Done when:** Either route handlers look up coordinator at request time via a mutable closure, or the server is torn down and recreated when a coordinator registers.
25-
26-
---
27-
2812
## Known edge cases — no fix yet
2913

3014
### 7. Autofire expiry window — coordinator in long tool call during countdown

0 commit comments

Comments
 (0)