Opus 4.6 and Codex 5.3 updates #874

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged

rblalock merged 6 commits into main from opus46-codex53

Feb 5, 2026

packages/opencode/AGENTS.md

-Original file line number
+Diff line change
@@ Expand Up / @@ -70,7 +70,6 @@ src/ @@
     │       └── entities.ts   # Entity extraction and management
     ├── tools/
     │   ├── index.ts          # Tool exports
-    │   ├── delegate.ts       # agentuity_coder_delegate tool
     │   └── background.ts     # Background task tools
     ├── background/
     │   ├── index.ts          # Background task exports
@@ Expand Down Expand Up @@
     ## Delegation Pattern
-    Lead delegates to specialized agents via `agentuity_coder_delegate` tool:
+    Lead delegates to specialized agents via two mechanisms:
-    ```typescript
-    {
-      agent: 'scout' | 'builder' | 'architect' | 'reviewer' | 'memory' | 'reasoner' | 'expert' | 'runner' | 'product' | 'monitor',
-      task: 'Description of what to do',
-      context?: 'Additional context'
-    }
-    ```
+    - **Task tool** (blocking) — Spawns a subagent, waits for result. Use for sequential work.
+    - **`agentuity_background_task`** (parallel) — Launches a task in a separate session. Use for independent concurrent work.
     ## Testing
@@ Expand Down @@

packages/opencode/src/agents/architect.ts

-Original file line number
+Diff line change
@@ Expand Up / @@ -259,7 +259,7 @@ export const architectAgent: AgentDefinition = { @@
     	displayName: 'Agentuity Coder Architect',
     	description:
     		'Senior implementer for complex autonomous tasks - Cadence mode, deep reasoning, extended execution',
-    	defaultModel: 'openai/gpt-5.2-codex',
+    	defaultModel: 'openai/gpt-5.3-codex',
     	systemPrompt: ARCHITECT_SYSTEM_PROMPT,
     	reasoningEffort: 'xhigh', // Maximum reasoning for complex tasks
     	temperature: 0.1, // Deterministic - precise code generation
@@ Expand Down @@

packages/opencode/src/agents/builder.ts

-Original file line number
+Diff line change
@@ Expand Up / @@ -430,7 +430,7 @@ export const builderAgent: AgentDefinition = { @@
     	id: 'ag-builder',
     	displayName: 'Agentuity Coder Builder',
     	description: 'Agentuity Coder implementer - writes code, makes edits, runs tests and builds',
-    	defaultModel: 'anthropic/claude-opus-4-5-20251101',
+    	defaultModel: 'anthropic/claude-opus-4-6',
     	systemPrompt: BUILDER_SYSTEM_PROMPT,
     	variant: 'high', // Careful thinking for implementation
     	temperature: 0.1, // Deterministic - precise code generation
@@ Expand Down @@

packages/opencode/src/agents/lead.ts

-Original file line number
+Diff line change
@@ Expand Up / @@ -1334,7 +1334,7 @@ export const leadAgent: AgentDefinition = { @@
     	displayName: 'Agentuity Coder Lead',
     	description:
     		'Agentuity Coder team orchestrator - delegates to Scout, Builder, Reviewer, Memory, Expert',
-    	defaultModel: 'anthropic/claude-opus-4-5-20251101',
+    	defaultModel: 'anthropic/claude-opus-4-6',
     	systemPrompt: LEAD_SYSTEM_PROMPT,
     	mode: 'all',
     	tools: {
@@ Expand Down @@

packages/opencode/src/agents/memory.ts

            
                      Original file line number
                      Diff line number
                      Diff line change
                  
    @@ -147,8 +147,8 @@ Agents can have different views of each other. Store and retrieve perspectives t
  
      "perspectiveId": "lead:view:builder",

      "observer": "entity:agent:lead",

      "observed": "entity:agent:builder",

      "observerModel": "claude-opus-4-5-20251101",

      "observedModel": "claude-opus-4-5-20251101",

      "observerModel": "claude-opus-4-6",

      "observedModel": "claude-opus-4-6",

      "conclusions": [

        {

          "type": "inductive",

    @@ -176,8 +176,8 @@ agentuity cloud kv set agentuity-opencode-memory "perspective:lead:builder" '{
  
      "perspectiveId": "lead:view:builder",

      "observer": "entity:agent:lead",

      "observed": "entity:agent:builder",

      "observerModel": "claude-opus-4-5-20251101",

      "observedModel": "claude-opus-4-5-20251101",

      "observerModel": "claude-opus-4-6-20260205",

      "observedModel": "claude-opus-4-6-20260205",

      "conclusions": [...],

      "recommendations": [...],

      "createdAt": "...",

packages/opencode/src/agents/monitor.ts

-Original file line number
+Diff line change
@@ Expand Up / @@ -99,7 +99,6 @@ export const monitorAgent: AgentDefinition = { @@
     			'glob',
     			'grep',
     			'task',
-    			'agentuity_coder_delegate',
     			'agentuity_background_task',
     			'agentuity_background_cancel',
     			'agentuity_memory_share',
@@ Expand Down @@

packages/opencode/src/agents/reasoner.ts

-Original file line number
+Diff line change
@@ Expand Up @@
     ### How to Query
-    Use \`agentuity_coder_delegate\` to ask Memory:
+    Use the Task tool to ask Memory:
     \`\`\`
-    agentuity_coder_delegate({
-      agent: "memory",
-      task: "What auth patterns and corrections do we have?",
-      context: "Reasoning about auth implementation in session data"
-    })
+    @Agentuity Coder Memory
+    What auth patterns and corrections do we have? Context: Reasoning about auth implementation in session data.
     \`\`\`
     ### The Feedback Loop
@@ Expand Down Expand Up / @@ -260,7 +258,7 @@ export const reasonerAgent: AgentDefinition = { @@
     	systemPrompt: REASONER_SYSTEM_PROMPT,
     	mode: 'subagent',
     	tools: {
-    		exclude: ['write', 'edit', 'apply_patch', 'task'],
+    		exclude: ['write', 'edit', 'apply_patch'],
     	},
     	reasoningEffort: 'high',
     	temperature: 0.3,
@@ Expand Down @@

packages/opencode/src/background/manager.ts

-Original file line number
+Diff line change
@@ Expand Up / @@ -568,6 +568,29 @@ export class BackgroundManager { @@
     	private async notifyParent(task: BackgroundTask): Promise<void> {
     		if (!task.parentSessionId) return;
+    		// Prevent duplicate notifications for the same task+status combination
+    		// This guards against OpenCode firing multiple events for the same status transition
+    		const notifiedStatuses = task.notifiedStatuses ?? new Set();
+    		// Self-healing for tasks created before deduplication was added:
+    		// If a task is already in a terminal state but has no notification history,
+    		// assume it was already notified and skip to prevent duplicate notifications.
+    		if (
+    			notifiedStatuses.size === 0 &&
+    			(task.status === 'completed' || task.status === 'error' || task.status === 'cancelled')
+    		) {
+    			notifiedStatuses.add(task.status);
+    			task.notifiedStatuses = notifiedStatuses;
+    			return;
+    		}
+    		if (notifiedStatuses.has(task.status)) {
+    			return; // Already notified for this status, skip duplicate
+    		}
+    		// Mark as notified BEFORE sending to prevent race conditions
+    		notifiedStatuses.add(task.status);
+    		task.notifiedStatuses = notifiedStatuses;
     		const statusLine = task.status === 'completed' ? 'completed' : task.status;
     		const message = `[BACKGROUND TASK ${statusLine.toUpperCase()}]
@@ Expand Down @@

packages/opencode/src/background/types.ts

-Original file line number
+Diff line change
@@ Expand Up / @@ -26,6 +26,7 @@ export interface BackgroundTask { @@
     	progress?: TaskProgress;
     	concurrencyKey?: string; // Active concurrency slot key
     	concurrencyGroup?: string; // Persistent key for re-acquiring on resume
+    	notifiedStatuses?: Set<BackgroundTaskStatus>; // Tracks statuses already notified to prevent duplicates
     }
     export interface LaunchInput {
@@ Expand Down @@

packages/opencode/src/config/loader.ts

-Original file line number
+Diff line change
@@ Expand Up / @@ -74,7 +74,7 @@ export async function getConfigPath(): Promise<string> { @@
      * {
      *   "agent": {
      *     "Agentuity Coder Architect": {
-     *       "model": "openai/gpt-5.2-codex",
+     *       "model": "openai/gpt-5.3-codex",
      *       "reasoningEffort": "xhigh"
      *     }
      *   }
@@ Expand Down @@

packages/opencode/src/plugin/plugin.ts

-Original file line number
+Diff line change
@@ Expand Up @@
     // Agents that should receive sandbox context in their prompts
     const SANDBOX_AWARE_AGENTS: AgentRole[] = ['lead', 'builder', 'architect'];
-    // Agent display names for @mentions
-    // Note: Monitor and Expert sub-agents have hidden: true so they won't appear in @ autocomplete,
-    // but they're still included here for programmatic invocation via Task tool
-    const AGENT_MENTIONS: Record<AgentRole, string> = {
-    	lead: '@Agentuity Coder Lead',
-    	scout: '@Agentuity Coder Scout',
-    	builder: '@Agentuity Coder Builder',
-    	architect: '@Agentuity Coder Architect',
-    	reviewer: '@Agentuity Coder Reviewer',
-    	memory: '@Agentuity Coder Memory',
-    	expert: '@Agentuity Coder Expert',
-    	'expert-backend': '@Agentuity Coder Expert Backend',
-    	'expert-frontend': '@Agentuity Coder Expert Frontend',
-    	'expert-ops': '@Agentuity Coder Expert Ops',
-    	runner: '@Agentuity Coder Runner',
-    	reasoner: '@Agentuity Coder Reasoner',
-    	product: '@Agentuity Coder Product',
-    	monitor: '@Agentuity Coder Monitor',
-    };
     export async function createCoderPlugin(ctx: PluginInput): Promise<Hooks> {
     	ctx.client.app.log({
     		body: {
@@ Expand Down Expand Up @@
     	// Use the schema from @opencode-ai/plugin's tool helper to avoid Zod version mismatches
     	const s = tool.schema;
-    	const coderDelegate = tool({
-    		description: `Delegate a task to a specialized Agentuity Coder agent.
-    Use this to:
-    - Scout: Explore codebase, find patterns, research documentation
-    - Builder: Implement features, write code, run tests (interactive work)
-    - Architect: Complex autonomous tasks, Cadence mode, deep reasoning (GPT Codex)
-    - Reviewer: Review changes, catch issues, apply fixes
-    - Memory: Store context, remember decisions across sessions
-    - Reasoner: Extract structured conclusions, resolve conflicts, surface corrections
-    - Expert: Get help with Agentuity CLI and cloud services
-    - Runner: Execute lint/build/test/typecheck/format commands, returns structured results
-    - Monitor: Watch background tasks and report when they complete`,
-    		args: {
-    			agent: s
-    				.enum([
-    					'scout',
-    					'builder',
-    					'architect',
-    					'reviewer',
-    					'memory',
-    					'reasoner',
-    					'expert',
-    					'runner',
-    					'product',
-    					'monitor',
-    				])
-    				.describe('Which agent to delegate to'),
-    			task: s.string().describe('Clear description of the task'),
-    			context: s.string().optional().describe('Additional context from previous tasks'),
-    		},
-    		async execute(args) {
-    			const mention = AGENT_MENTIONS[args.agent as AgentRole];
-    			let prompt = `${mention}\n\n## Task\n${args.task}`;
-    			if (args.context) {
-    				prompt = `${mention}\n\n## Context\n${args.context}\n\n## Task\n${args.task}`;
-    			}
-    			return `To delegate this task, use the Task tool with this prompt:\n\n${prompt}\n\nThe ${args.agent} agent will handle this task.`;
-    		},
-    	});
     	const backgroundTask = tool({
     		description: `Launch a task to run in the background. Use this for parallel execution of multiple independent tasks.
@@ Expand Down Expand Up @@
     	});
     	return {
-    		agentuity_coder_delegate: coderDelegate,
     		agentuity_background_task: backgroundTask,
     		agentuity_background_output: backgroundOutput,
     		agentuity_background_cancel: backgroundCancel,
@@ Expand Down @@

packages/opencode/src/tools/delegate.ts

This file was deleted.

packages/opencode/src/tools/index.ts

-Original file line number
+Diff line change
@@ -1,4 +1,3 @@
-    export { delegateTool, DelegateArgsSchema, type DelegateArgs } from './delegate';
     export {
     	createBackgroundTools,
     	BackgroundTaskArgsSchema,
@@ Expand Down @@

packages/opencode/test/agents.test.ts

            
                      Original file line number
                      Diff line number
                      Diff line change
                  
    @@ -104,7 +104,7 @@ describe('Agents', () => {
  
    		it('Architect agent has GPT Codex with xhigh reasoning', () => {

    			const architect = agents.architect;

    			expect(architect.defaultModel).toBe('openai/gpt-5.2-codex');

    			expect(architect.defaultModel).toBe('openai/gpt-5.3-codex');

    			expect(architect.reasoningEffort).toBe('xhigh');

    			expect(architect.temperature).toBe(0.1);

    			expect(architect.systemPrompt).toContain('Cadence');

    @@ -149,7 +149,7 @@ describe('Agents', () => {
  
    			expect(agents.reasoner.mode).toBe('subagent');

    			expect(agents.reasoner.tools?.exclude).toContain('write');

    			expect(agents.reasoner.tools?.exclude).toContain('edit');

    			expect(agents.reasoner.tools?.exclude).toContain('task');

    			expect(agents.reasoner.tools?.exclude).not.toContain('task');

    			expect(agents.reasoner.defaultModel).toBe('openai/gpt-5.2');

    			expect(agents.reasoner.temperature).toBe(0.3);

    		});

packages/opencode/test/config.test.ts

            
                      Original file line number
                      Diff line number
                      Diff line change
                  
    @@ -54,12 +54,12 @@ describe('Config', () => {
  
    	describe('Model helpers', () => {

    		it('isOpenAIModel correctly identifies OpenAI models', () => {

    			expect(isOpenAIModel('openai/gpt-5.2')).toBe(true);

    			expect(isOpenAIModel('openai/gpt-5.2-codex')).toBe(true);

    			expect(isOpenAIModel('anthropic/claude-opus-4-5')).toBe(false);

    			expect(isOpenAIModel('openai/gpt-5.3-codex')).toBe(true);

    			expect(isOpenAIModel('anthropic/claude-opus-4-6')).toBe(false);

    		});

    		it('isAnthropicModel correctly identifies Anthropic models', () => {

    			expect(isAnthropicModel('anthropic/claude-opus-4-5')).toBe(true);

    			expect(isAnthropicModel('anthropic/claude-opus-4-6')).toBe(true);

    			expect(isAnthropicModel('anthropic/claude-haiku-4-5')).toBe(true);

    			expect(isAnthropicModel('openai/gpt-5.2')).toBe(false);

    		});

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Opus 4.6 and Codex 5.3 updates #874

Uh oh!

Diff view

Diff view

There are no files selected for viewing

Uh oh!

Uh oh!

Uh oh!