feat(cli): parallel subagents

[uinstinct]

Description

add parallel subagents

requires #10830 to merged first

AI Code Review

• Team members only: AI review runs automatically when PR is opened or marked ready for review
• Team members can also trigger a review by commenting @continue-review

Checklist

• [] I've read the contributing guide
• [] The relevant docs, if any, have been updated or created
• [] The relevant tests, if any, have been updated or created

Screen recording or screenshot

[ When applicable, please include a short screen recording or screenshot - this makes it much easier for us as contributors to review and understand your changes. See this PR as a good example. ]

Tests

[ What tests were added or updated to ensure the changes work as expected? ]

---

Continue Tasks: ❌ 7 failed — View all

---

Summary by cubic

Enables parallel subagent execution in the CLI with isolated per-run context and live streaming output. Adds a parallel task runner and built-in Navigator and Generalist subagents, and shows each subagent’s output live (including parallel runs).

• New Features

  • SubAgentService scopes system messages and tool permissions per execution, streams per-subagent events, and blocks nested subagent calls.
  • SubagentParallel tool runs multiple subagents concurrently and returns aggregated results.
  • Built-in subagents: Navigator and Generalist; auto-added when using Anthropic with direct API keys.
  • SubAgentOutput TUI renders live, separate output for each subagent, including parallel runs; tool results remain separate.

• Refactors

  • Replaced executor with SubAgentService, registered in the service container; unified config with built-in subagents and per-subagent LLM APIs; refined subagent tool description.
  • Removed the subagent beta flag and CLI option; tools are enabled by default. When ChatHistoryService is disabled, tool results/status are written to local chat history.

^{Written for commit 698e237. Summary will update on new commits.}

[cubic-dev-ai]

4 issues found across 20 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="extensions/cli/src/services/SubAgentService.ts">

<violation number="1" location="extensions/cli/src/services/SubAgentService.ts:127">
P2: Parallel subagent executions race on the global chatHistorySvc.isReady override, so completion order can restore a stale override and leave chat history disabled or re-enable it while another subagent still runs.</violation>
</file>

<file name="extensions/cli/src/tools/subagent.test.ts">

<violation number="1" location="extensions/cli/src/tools/subagent.test.ts:5">
P2: The updated import now points to `getAgents.js`, but the test still mocks `get-agents.js`, so Vitest won’t apply the mock and the test will call real implementations. Update the mock path to match the import specifier.</violation>
</file>

<file name="extensions/cli/src/ui/components/SubAgentOutput.tsx">

<violation number="1" location="extensions/cli/src/ui/components/SubAgentOutput.tsx:57">
P2: `contentLines` grows without bound even though only the last 15 lines are rendered, which can lead to unnecessary memory usage and O(n) updates for long-running or chatty subagents.</violation>
</file>

<file name="extensions/cli/src/ui/ToolResultSummary.tsx">

<violation number="1" location="extensions/cli/src/ui/ToolResultSummary.tsx:119">
P2: Removed Subagent-specific summary handling while Subagent outputs still include `<task_metadata>` tags. The generic summary now surfaces raw metadata (or only line/char counts) and loses the prior “Subagent executing…” state for metadata-only updates, regressing the user-facing CLI summary.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

P2: Parallel subagent executions race on the global chatHistorySvc.isReady override, so completion order can restore a stale override and leave chat history disabled or re-enable it while another subagent still runs.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At extensions/cli/src/services/SubAgentService.ts, line 127:

<comment>Parallel subagent executions race on the global chatHistorySvc.isReady override, so completion order can restore a stale override and leave chat history disabled or re-enable it while another subagent still runs.</comment>

<file context>
@@ -0,0 +1,277 @@
+          : undefined;
+
+      if (chatHistorySvc && originalIsReady) {
+        chatHistorySvc.isReady = () => false;
+      }
+
</file context>

[cubic-dev-ai]

P2: The updated import now points to getAgents.js, but the test still mocks get-agents.js, so Vitest won’t apply the mock and the test will call real implementations. Update the mock path to match the import specifier.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At extensions/cli/src/tools/subagent.test.ts, line 5:

<comment>The updated import now points to `getAgents.js`, but the test still mocks `get-agents.js`, so Vitest won’t apply the mock and the test will call real implementations. Update the mock path to match the import specifier.</comment>

<file context>
@@ -1,14 +1,17 @@
-import { executeSubAgent } from "../subagent/executor.js";
-import { getAgentNames, getSubagent } from "../subagent/get-agents.js";
+import { subAgentService } from "../services/SubAgentService.js";
+import { getAgentNames, getSubagent } from "../subagent/getAgents.js";
 
 import { subagentTool } from "./subagent.js";
</file context>

[cubic-dev-ai]

P2: contentLines grows without bound even though only the last 15 lines are rendered, which can lead to unnecessary memory usage and O(n) updates for long-running or chatty subagents.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At extensions/cli/src/ui/components/SubAgentOutput.tsx, line 57:

<comment>`contentLines` grows without bound even though only the last 15 lines are rendered, which can lead to unnecessary memory usage and O(n) updates for long-running or chatty subagents.</comment>

<file context>
@@ -0,0 +1,132 @@
+          ...prev,
+          [data.executionId]: {
+            ...agent,
+            contentLines: updatedLines,
+          },
+        };
</file context>

[cubic-dev-ai]

P2: Removed Subagent-specific summary handling while Subagent outputs still include <task_metadata> tags. The generic summary now surfaces raw metadata (or only line/char counts) and loses the prior “Subagent executing…” state for metadata-only updates, regressing the user-facing CLI summary.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At extensions/cli/src/ui/ToolResultSummary.tsx, line 119:

<comment>Removed Subagent-specific summary handling while Subagent outputs still include `<task_metadata>` tags. The generic summary now surfaces raw metadata (or only line/char counts) and loses the prior “Subagent executing…” state for metadata-only updates, regressing the user-facing CLI summary.</comment>

<file context>
@@ -115,57 +115,6 @@ const ToolResultSummary: React.FC<ToolResultSummaryProps> = ({
-  }
-
   // Handle all other cases with text summary
   const getSummary = () => {
     // Check if this is a user cancellation first
</file context>