@agentuity/opencode 1.0.14 → 1.0.16

package/dist/types.js.map

@@ -1 +1 @@
-{"version":3,"file":"types.js","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAiBxB,MAAM,CAAC,MAAM,eAAe,GAAG,CAAC,CAAC,IAAI,CAAC;IACrC,MAAM;IACN,OAAO;IACP,SAAS;IACT,WAAW;IACX,UAAU;IACV,QAAQ;IACR,QAAQ;IACR,gBAAgB;IAChB,iBAAiB;IACjB,YAAY;IACZ,QAAQ;IACR,SAAS;IACT,SAAS;CACT,CAAC,CAAC;AAGH,MAAM,CAAC,MAAM,gBAAgB,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,SAAS,EAAE,SAAS,EAAE,WAAW,EAAE,QAAQ,EAAE,WAAW,CAAC,CAAC,CAAC;AAGnG,MAAM,CAAC,MAAM,0BAA0B,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,QAAQ,EAAE,QAAQ,EAAE,UAAU,CAAC,CAAC,CAAC;AAGnF,MAAM,CAAC,MAAM,mBAAmB,GAAG,CAAC,CAAC,IAAI,CAAC;IACzC,SAAS;IACT,QAAQ;IACR,WAAW;IACX,QAAQ;IACR,WAAW;CACX,CAAC,CAAC;AAGH,MAAM,CAAC,MAAM,wBAAwB,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,eAAe,EAAE,YAAY,CAAC,CAAC,CAAC;AAgEvF,2DAA2D;AAC3D,MAAM,CAAC,MAAM,oBAAoB,GAAG,CAAC,CAAC,MAAM,CAAC;IAC5C,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,SAAS,EAAE,UAAU,CAAC,CAAC;IACrC,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;CACnC,CAAC,CAAC;AAGH,yCAAyC;AACzC,MAAM,CAAC,MAAM,qBAAqB,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,QAAQ,EAAE,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC;AAGhF,kDAAkD;AAClD,MAAM,CAAC,MAAM,kBAAkB,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,QAAQ,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,CAAC;AAmB3E,MAAM,CAAC,MAAM,sBAAsB,GAAG,CAAC,CAAC,MAAM,CAAC;IAC9C,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAC5B,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,EAAE;IAChD,OAAO,EAAE,kBAAkB,CAAC,QAAQ,EAAE;IACtC,eAAe,EAAE,qBAAqB,CAAC,QAAQ,EAAE;IACjD,QAAQ,EAAE,oBAAoB,CAAC,QAAQ,EAAE;IACzC,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;CAC/B,CAAC,CAAC;AAYH,MAAM,CAAC,MAAM,0BAA0B,GAAG,CAAC,CAAC,MAAM,CAAC;IAClD,OAAO,EAAE,CAAC,CAAC,OAAO,EAAE;IACpB,kBAAkB,EAAE,CAAC,CAAC,MAAM,EAAE;IAC9B,cAAc,EAAE,CAAC,CAAC,MAAM,EAAE;IAC1B,mBAAmB,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;IAChE,gBAAgB,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;CAC7D,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,kBAAkB,GAAG,CAAC,CAAC,MAAM,CAAC;IAC1C,OAAO,EAAE,CAAC,CAAC,OAAO,EAAE;IACpB,KAAK,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;IACrC,QAAQ,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;CACxC,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,gBAAgB,GAAG,CAAC,CAAC,MAAM,CAAC;IACxC,OAAO,EAAE,CAAC,CAAC,OAAO,EAAE;IACpB,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE;IACpB,gBAAgB,EAAE,CAAC,CAAC,MAAM,EAAE;IAC5B,iBAAiB,EAAE,CAAC,CAAC,MAAM,EAAE;CAC7B,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;IACzC,GAAG,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAC1B,YAAY,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;IAC5C,eAAe,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;IAC/C,UAAU,EAAE,0BAA0B,CAAC,QAAQ,EAAE;IACjD,MAAM,EAAE,kBAAkB,CAAC,QAAQ,EAAE;IACrC,IAAI,EAAE,gBAAgB,CAAC,QAAQ,EAAE;CACjC,CAAC,CAAC;AAuBH,+DAA+D"}
+{"version":3,"file":"types.js","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAiBxB,MAAM,CAAC,MAAM,eAAe,GAAG,CAAC,CAAC,IAAI,CAAC;IACrC,MAAM;IACN,OAAO;IACP,SAAS;IACT,WAAW;IACX,UAAU;IACV,QAAQ;IACR,QAAQ;IACR,gBAAgB;IAChB,iBAAiB;IACjB,YAAY;IACZ,QAAQ;IACR,SAAS;IACT,SAAS;CACT,CAAC,CAAC;AAGH,MAAM,CAAC,MAAM,gBAAgB,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,SAAS,EAAE,SAAS,EAAE,WAAW,EAAE,QAAQ,EAAE,WAAW,CAAC,CAAC,CAAC;AAGnG,MAAM,CAAC,MAAM,0BAA0B,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,QAAQ,EAAE,QAAQ,EAAE,UAAU,CAAC,CAAC,CAAC;AAGnF,MAAM,CAAC,MAAM,mBAAmB,GAAG,CAAC,CAAC,IAAI,CAAC;IACzC,SAAS;IACT,QAAQ;IACR,WAAW;IACX,QAAQ;IACR,WAAW;CACX,CAAC,CAAC;AAGH,MAAM,CAAC,MAAM,wBAAwB,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,eAAe,EAAE,YAAY,CAAC,CAAC,CAAC;AAgEvF,2DAA2D;AAC3D,MAAM,CAAC,MAAM,oBAAoB,GAAG,CAAC,CAAC,MAAM,CAAC;IAC5C,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,SAAS,EAAE,UAAU,CAAC,CAAC;IACrC,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;CACnC,CAAC,CAAC;AAGH,yCAAyC;AACzC,MAAM,CAAC,MAAM,qBAAqB,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,QAAQ,EAAE,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC;AAGhF,kDAAkD;AAClD,MAAM,CAAC,MAAM,kBAAkB,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,QAAQ,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,CAAC;AAmB3E,MAAM,CAAC,MAAM,sBAAsB,GAAG,CAAC,CAAC,MAAM,CAAC;IAC9C,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAC5B,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,EAAE;IAChD,OAAO,EAAE,kBAAkB,CAAC,QAAQ,EAAE;IACtC,eAAe,EAAE,qBAAqB,CAAC,QAAQ,EAAE;IACjD,QAAQ,EAAE,oBAAoB,CAAC,QAAQ,EAAE;IACzC,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;CAC/B,CAAC,CAAC;AA+BH,MAAM,CAAC,MAAM,0BAA0B,GAAG,CAAC,CAAC,MAAM,CAAC;IAClD,OAAO,EAAE,CAAC,CAAC,OAAO,EAAE;IACpB,kBAAkB,EAAE,CAAC,CAAC,MAAM,EAAE;IAC9B,cAAc,EAAE,CAAC,CAAC,MAAM,EAAE;IAC1B,mBAAmB,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;IAChE,gBAAgB,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;CAC7D,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,kBAAkB,GAAG,CAAC,CAAC,MAAM,CAAC;IAC1C,OAAO,EAAE,CAAC,CAAC,OAAO,EAAE;IACpB,KAAK,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;IACrC,QAAQ,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;CACxC,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,gBAAgB,GAAG,CAAC,CAAC,MAAM,CAAC;IACxC,OAAO,EAAE,CAAC,CAAC,OAAO,EAAE;IACpB,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE;IACpB,gBAAgB,EAAE,CAAC,CAAC,MAAM,EAAE;IAC5B,iBAAiB,EAAE,CAAC,CAAC,MAAM,EAAE;CAC7B,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,sBAAsB,GAAG,CAAC,CAAC,MAAM,CAAC;IAC9C,YAAY,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;IACpC,cAAc,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;IACtC,cAAc,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;IACtC,oBAAoB,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAC3C,YAAY,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;IACpC,gBAAgB,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IACvC,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;CAC/B,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;IACzC,GAAG,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAC1B,YAAY,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;IAC5C,eAAe,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,QAAQ,EAAE;IAC/C,UAAU,EAAE,0BAA0B,CAAC,QAAQ,EAAE;IACjD,MAAM,EAAE,kBAAkB,CAAC,QAAQ,EAAE;IACrC,IAAI,EAAE,gBAAgB,CAAC,QAAQ,EAAE;IACjC,UAAU,EAAE,sBAAsB,CAAC,QAAQ,EAAE;CAC7C,CAAC,CAAC;AAuBH,+DAA+D"}

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@agentuity/opencode",
-	"version": "1.0.14",
+	"version": "1.0.16",
 	"license": "Apache-2.0",
 	"author": "Agentuity employees and contributors",
 	"description": "Agentuity Open Code plugin with specialized AI coding agents",
@@ -40,13 +40,13 @@
 		"prepublishOnly": "bun run clean && bun run build"
 	},
 	"dependencies": {
-		"@agentuity/core": "1.0.14",
+		"@agentuity/core": "1.0.16",
 		"@opencode-ai/plugin": "^1.1.36",
 		"yaml": "^2.8.1",
 		"zod": "^4.3.5"
 	},
 	"devDependencies": {
-		"@agentuity/test-utils": "1.0.14",
+		"@agentuity/test-utils": "1.0.16",
 		"@types/bun": "latest",
 		"bun-types": "latest",
 		"typescript": "^5.9.0"

package/src/agents/expert-backend.ts

@@ -487,7 +487,7 @@ export const expertBackendAgent: AgentDefinition = {
 	id: 'ag-expert-backend',
 	displayName: 'Agentuity Coder Expert Backend',
 	description: 'Agentuity backend specialist - runtime, agents, schemas, drizzle, postgres, evals',
-	defaultModel: 'anthropic/claude-sonnet-4-5-20250929',
+	defaultModel: 'anthropic/claude-sonnet-4-6',
 	systemPrompt: EXPERT_BACKEND_SYSTEM_PROMPT,
 	mode: 'subagent',
 	hidden: true, // Only invoked by Expert orchestrator

package/src/agents/expert-frontend.ts

@@ -474,7 +474,7 @@ export const expertFrontendAgent: AgentDefinition = {
 	id: 'ag-expert-frontend',
 	displayName: 'Agentuity Coder Expert Frontend',
 	description: 'Agentuity frontend specialist - React hooks, auth, workbench, web utilities',
-	defaultModel: 'anthropic/claude-sonnet-4-5-20250929',
+	defaultModel: 'anthropic/claude-sonnet-4-6',
 	systemPrompt: EXPERT_FRONTEND_SYSTEM_PROMPT,
 	mode: 'subagent',
 	hidden: true, // Only invoked by Expert orchestrator

package/src/agents/expert-ops.ts

@@ -369,7 +369,7 @@ export const expertOpsAgent: AgentDefinition = {
 	id: 'ag-expert-ops',
 	displayName: 'Agentuity Coder Expert Ops',
 	description: 'Agentuity operations specialist - CLI, cloud services, deployments, sandboxes',
-	defaultModel: 'anthropic/claude-sonnet-4-5-20250929',
+	defaultModel: 'anthropic/claude-sonnet-4-6',
 	systemPrompt: EXPERT_OPS_SYSTEM_PROMPT,
 	mode: 'subagent',
 	hidden: true, // Only invoked by Expert orchestrator

package/src/agents/expert.ts

@@ -214,7 +214,7 @@ export const expertAgent: AgentDefinition = {
 	id: 'ag-expert',
 	displayName: 'Agentuity Coder Expert',
 	description: 'Agentuity Coder Agentuity specialist - knows CLI, SDK, cloud services deeply',
-	defaultModel: 'anthropic/claude-sonnet-4-5-20250929',
+	defaultModel: 'anthropic/claude-sonnet-4-6',
 	systemPrompt: EXPERT_SYSTEM_PROMPT,
 	variant: 'high', // Careful thinking for technical guidance
 	temperature: 0.1, // Accurate, consistent technical answers

package/src/agents/lead.ts

@@ -14,7 +14,7 @@ You are the Lead agent on the Agentuity Coder team — the **air traffic control
 | Quality gatekeeper             | Cloud operator                 |
 | Context coordinator            | Test runner                    |
 
-**Golden Rule**: If it involves writing code, editing files, running commands, or searching codebases — delegate it. Your job is to think, plan, coordinate, and decide.
+**Golden Rule**: If it involves writing code, editing files, running commands, searching codebases, or gathering information via research — default to delegating it. Your job is to think, plan, coordinate, and decide. You CAN do lightweight research when working solo on simple tasks, but once you've delegated work to background agents, commit fully to the orchestration role.
 
 ## Delegation Decision Guide
 
@@ -663,17 +663,44 @@ When you have launched background tasks via \`agentuity_background_task\`:
 
 **The whole point of background tasks is parallel execution by OTHER agents.** If you do the work yourself while they're running, you waste tokens and create conflicting results.
 
-**What you CAN do while waiting:**
-- Work on DIFFERENT, unrelated tasks
-- Plan next steps for when results arrive
+### Tool Restrictions While Background Tasks Are Running
+
+Once you have launched background tasks, you enter **orchestration-only mode**. Do NOT use research or exploration tools until background tasks have returned.
+
+**Tools you MUST NOT use while background tasks are pending:**
+- \`webfetch\` — do not fetch any URLs (even "different" ones related to the task)
+- \`grep\` / \`glob\` — do not search the codebase for research
+- \`read\` — do not read source files for research (reading task state or config is OK)
+- \`bash\` — do not run exploratory commands
+
+**What you CAN do while waiting (exhaustive list):**
+- Poll background task status with \`agentuity_background_output\` or \`agentuity_background_inspect\`
 - Answer user questions about progress
-- Update task state in KV
+- Update the todo list
+- Use extended thinking to reason about how you'll combine results (no tool calls — just think)
 
 **What you MUST NOT do:**
-- Start doing the same work you delegated
-- "Get impatient" and bypass the background agents
+- Use ANY research tool — if you catch yourself reaching for webfetch, grep, glob, or read to "get a head start" or "do something useful while waiting," STOP. That IS the background agents' job.
+- Rationalize research as "planning" — planning while waiting means thinking, not fetching or searching
+- Start "different but related" research — if the background tasks are researching a feature, do not research adjacent aspects of that feature yourself
 - Assume background tasks failed just because they haven't returned yet
 
+## Context Budget Awareness
+
+Your context window is finite and shared between everything you do. Every tool call output — especially \`webfetch\` responses and file reads — consumes context that you need later for:
+- Processing background task results when they return
+- Synthesizing information from multiple agents
+- Making strategic decisions with full awareness
+
+**A single webfetch response can consume 5-15% of your context.** Three unnecessary fetches while waiting for background tasks can waste 30-45% of your context — potentially leaving you unable to properly process the actual results you delegated for.
+
+**Before using any research tool, ask yourself:**
+1. "Is a background agent already getting this information?" → If yes, WAIT.
+2. "Do I need this to make a decision RIGHT NOW?" → If no, WAIT.
+3. "Will this output be large?" → If yes, delegate it.
+
+When in doubt, preserve your context. You need it most when results start flowing back from your agents.
+
 ## Task Completion: Memorialize the Session
 
 **IMPORTANT:** When you complete a task, ALWAYS tell Memory to save the session to vector storage.

package/src/agents/monitor.ts

@@ -4,10 +4,17 @@ export const MONITOR_SYSTEM_PROMPT = `# BackgroundMonitor Agent
 
 You are a background task monitor. Your ONLY job is to watch background tasks and report when they complete.
 
+## Primary Notification Channel
+
+Background tasks automatically notify Lead with messages like:
+\`[BACKGROUND TASK COMPLETED]\`
+
+Those event-driven notifications are the primary mechanism. You are a fallback for Lead-of-Leads scenarios where multiple child Leads are running and a summary pass is needed.
+
 ## How You Work
 
 1. You receive a list of task IDs to monitor
-2. You poll their status using agentuity_background_output
+2. You check their status using agentuity_background_output
 3. When ALL tasks complete (or error), you report back to Lead
 4. You do NOT interpret results - just report completion status
 
@@ -20,36 +27,36 @@ When you need deeper insight into a task, use \`agentuity_background_inspect\` w
 - Cost summary (total cost + tokens)
 - Child session count (for nested Lead-of-Leads)
 
-Use inspect when a task has been running for many poll cycles without completing — it can reveal what the agent is stuck on.
+Use inspect when a task has been running for many check cycles without completing — it can reveal what the agent is stuck on.
 
 For a full session tree with all child sessions, costs, and health summary, use \`agentuity_session_dashboard({ session_id: "..." })\`. This is especially useful when monitoring Lead-of-Leads scenarios with multiple parallel workstreams.
 
-## Polling Behavior
+## Bounded Check Cycles
 
-- Poll every 10 seconds (wait between checks)
-- Continue until ALL tasks are complete or errored
-- No timeout - watch indefinitely
+- Run a short, bounded series of check cycles (e.g., 3–5 passes)
+- If tasks are still pending/running after the final pass, report the current status and highlight which tasks appear stuck
+- If tasks appear stuck, use \`agentuity_background_inspect\` for those tasks before reporting
 
-## Polling Process
+## Check Process
 
-For each poll cycle:
+For each check cycle:
 1. Check each task ID with \`agentuity_background_output({ task_id: "bg_xxx" })\`
 2. Track the status of each task
-3. If any task is still "pending" or "running", wait 10 seconds and poll again
-4. When all tasks are "completed" or "error", generate the final report
+3. If all tasks are "completed" or "error", generate the final report
+4. Otherwise, repeat for the next cycle (bounded)
 
 ## Report Format
 
-When all tasks complete, output:
+When all tasks complete (or when you finish the bounded cycles), output:
 
 \`\`\`markdown
-## Background Tasks Complete
+## Background Tasks Status
 
 | Task ID | Status | Summary |
 |---------|--------|---------|
 | bg_xxx | completed | [first 100 chars of result] |
 | bg_yyy | error | [error message] |
-| bg_zzz | completed | [first 100 chars of result] |
+| bg_zzz | running | [last known status] |
 
 ### Detailed Results
 
@@ -59,7 +66,7 @@ When all tasks complete, output:
 **bg_yyy (error):**
 [error message]
 
-All monitored tasks have finished. Lead can now proceed with integration.
+If any tasks are still running/pending after the final pass, list them under a short "Still Running" section and mention that Lead should wait for event-driven notifications or re-check later.
 \`\`\`
 
 ## What You Do NOT Do
@@ -69,27 +76,9 @@ All monitored tasks have finished. Lead can now proceed with integration.
 - ❌ Interact with the user
 - ❌ Modify any files
 - ❌ Call other agents
-- ❌ Use tools other than agentuity_background_output
+- ❌ Use tools other than agentuity_background_output, agentuity_background_inspect, and agentuity_session_dashboard
 
 You are a simple, focused watcher. Report completions, nothing more.
-
-## Example Workflow
-
-Given task: "Monitor these tasks: bg_abc123, bg_def456"
-
-1. Call agentuity_background_output for bg_abc123
-2. Call agentuity_background_output for bg_def456
-3. If any status is "pending" or "running", wait 10 seconds
-4. Repeat steps 1-3 until all complete
-5. Output final report
-
-## Waiting Between Polls
-
-Since you cannot use setTimeout, after checking all tasks and finding some still running, respond with something like:
-
-"Polling cycle complete. Tasks still running: [list]. Waiting 10 seconds before next poll..."
-
-Then immediately poll again. The conversation history serves as your "timer" - each response and check adds natural delay.
 `;
 
 export const monitorAgent: AgentDefinition = {

package/src/agents/reviewer.ts

@@ -363,7 +363,7 @@ export const reviewerAgent: AgentDefinition = {
 	id: 'ag-reviewer',
 	displayName: 'Agentuity Coder Reviewer',
 	description: 'Agentuity Coder reviewer - reviews code, catches issues, applies fixes',
-	defaultModel: 'anthropic/claude-sonnet-4-5-20250929',
+	defaultModel: 'anthropic/claude-sonnet-4-6',
 	systemPrompt: REVIEWER_SYSTEM_PROMPT,
 	variant: 'high', // Careful thinking for thorough review
 	temperature: 0.1, // Consistent, deterministic reviews

package/src/agents/scout.ts

@@ -316,10 +316,10 @@ export const scoutAgent: AgentDefinition = {
 	displayName: 'Agentuity Coder Scout',
 	description:
 		'Agentuity Coder explorer - analyzes codebases, finds patterns, researches docs (read-only)',
-	defaultModel: 'anthropic/claude-haiku-4-5-20251001',
+	defaultModel: 'anthropic/claude-sonnet-4-6',
 	systemPrompt: SCOUT_SYSTEM_PROMPT,
 	tools: {
-		exclude: ['write', 'edit', 'apply_patch', 'bash'],
+		exclude: ['write', 'edit', 'apply_patch'],
 	},
 	// Scout uses default variant (speed over depth) and low temp for factual exploration
 	temperature: 0.0,

package/src/background/manager.ts

@@ -14,7 +14,7 @@ import { ConcurrencyManager } from './concurrency';
 
 const DEFAULT_BACKGROUND_CONFIG: BackgroundTaskConfig = {
 	enabled: true,
-	defaultConcurrency: 1,
+	defaultConcurrency: 5,
 	staleTimeoutMs: 30 * 60 * 1000,
 };
 
@@ -48,12 +48,15 @@ export class BackgroundManager {
 	private concurrency: ConcurrencyManager;
 	private callbacks?: BackgroundManagerCallbacks;
 	private dbReader?: OpenCodeDBReader;
+	private serverUrl: string | undefined;
+	private authHeaders: Record<string, string> | undefined;
 	private tasks = new Map<string, BackgroundTask>();
 	private tasksByParent = new Map<string, Set<string>>();
 	private tasksBySession = new Map<string, string>();
 	private notifications = new Map<string, Set<string>>();
 	private toolCallIds = new Map<string, Set<string>>();
 	private shuttingDown = false;
+	private refreshIntervalId: ReturnType<typeof setInterval> | undefined;
 
 	constructor(
 		ctx: PluginInput,
@@ -69,6 +72,68 @@ export class BackgroundManager {
 		});
 		this.callbacks = callbacks;
 		this.dbReader = dbReader;
+		this.serverUrl = this.resolveServerUrl();
+		this.authHeaders = this.resolveAuthHeaders();
+
+		// Periodic safety net: refresh task statuses every 30s in case events are missed
+		this.refreshIntervalId = setInterval(() => {
+			if (this.shuttingDown) return;
+			const hasActive = Array.from(this.tasks.values()).some(
+				(t) => t.status === 'pending' || t.status === 'running'
+			);
+			if (hasActive) {
+				void this.refreshStatuses();
+			}
+		}, 30_000);
+	}
+
+	/**
+	 * Resolve the server URL from the plugin context.
+	 * Mirrors the defensive pattern used in the tmux manager to handle
+	 * sandbox environments where the client may not have a baseUrl configured.
+	 */
+	private resolveServerUrl(): string | undefined {
+		const ctx = this.ctx as unknown as {
+			serverUrl?: string | URL;
+			baseUrl?: string | URL;
+			client?: { baseUrl?: string | URL };
+		};
+		const serverUrl = ctx.serverUrl ?? ctx.baseUrl ?? ctx.client?.baseUrl;
+		if (!serverUrl) return undefined;
+		const urlStr = typeof serverUrl === 'string' ? serverUrl : serverUrl.toString();
+		// Strip trailing slash to prevent double-slash when SDK appends paths like /session
+		return urlStr.replace(/\/+$/, '');
+	}
+
+	/**
+	 * Resolve authentication headers from environment variables.
+	 *
+	 * Reads `OPENCODE_SERVER_USERNAME` and `OPENCODE_SERVER_PASSWORD` (set
+	 * automatically by the OpenCode server in sandbox environments) and
+	 * produces a Basic Auth header (`base64("username:password")`).
+	 *
+	 * In sandbox environments the SDK client's default auth may not carry over
+	 * when a per-call `baseUrl` override is provided, so we need to explicitly
+	 * attach these credentials for server-to-server requests.
+	 */
+	private resolveAuthHeaders(): Record<string, string> | undefined {
+		const username = process.env.OPENCODE_SERVER_USERNAME;
+		const password = process.env.OPENCODE_SERVER_PASSWORD;
+		if (!username || !password) return undefined;
+		const encoded = Buffer.from(username + ':' + password).toString('base64');
+		return { Authorization: `Basic ${encoded}` };
+	}
+
+	/**
+	 * Build the per-call client overrides (baseUrl + auth headers).
+	 * Spread this into every SDK client call so both the server URL and
+	 * authentication are correctly forwarded in sandbox environments.
+	 */
+	private getClientOverrides(): { baseUrl?: string; headers?: Record<string, string> } {
+		const overrides: { baseUrl?: string; headers?: Record<string, string> } = {};
+		if (this.serverUrl) overrides.baseUrl = this.serverUrl;
+		if (this.authHeaders) overrides.headers = this.authHeaders;
+		return overrides;
 	}
 
 	async launch(input: LaunchInput): Promise<BackgroundTask> {
@@ -82,6 +147,7 @@ export class BackgroundManager {
 			status: 'pending',
 			queuedAt: new Date(),
 			concurrencyGroup: this.getConcurrencyGroup(input.agent),
+			notifiedStatuses: new Set(),
 		};
 
 		this.tasks.set(task.id, task);
@@ -188,12 +254,14 @@ export class BackgroundManager {
 			const sessionResponse = await this.ctx.client.session.get({
 				path: { id: task.sessionId },
 				throwOnError: false,
+				...this.getClientOverrides(),
 			});
 
 			// Get messages from the session
 			const messagesResponse = await this.ctx.client.session.messages({
 				path: { id: task.sessionId },
 				throwOnError: false,
+				...this.getClientOverrides(),
 			});
 
 			const session = unwrapResponse<unknown>(sessionResponse);
@@ -243,6 +311,7 @@ export class BackgroundManager {
 				const childrenResponse = await this.ctx.client.session.children({
 					path: { id: parentId },
 					throwOnError: false,
+					...this.getClientOverrides(),
 				});
 
 				const rawChildren = unwrapResponse<Array<unknown>>(childrenResponse);
@@ -334,6 +403,15 @@ export class BackgroundManager {
 								},
 							};
 
+							// Mark recovered terminal tasks as already notified
+							if (
+								task.status === 'completed' ||
+								task.status === 'error' ||
+								task.status === 'cancelled'
+							) {
+								task.notifiedStatuses = new Set([task.status]);
+							}
+
 							this.tasks.set(task.id, task);
 							this.tasksBySession.set(sess.id, task.id);
 
@@ -356,6 +434,7 @@ export class BackgroundManager {
 			// Get all sessions
 			const sessionsResponse = await this.ctx.client.session.list({
 				throwOnError: false,
+				...this.getClientOverrides(),
 			});
 
 			const rawSessions = unwrapResponse<Array<unknown>>(sessionsResponse);
@@ -409,6 +488,15 @@ export class BackgroundManager {
 						},
 					};
 
+					// Mark recovered terminal tasks as already notified
+					if (
+						task.status === 'completed' ||
+						task.status === 'error' ||
+						task.status === 'cancelled'
+					) {
+						task.notifiedStatuses = new Set([task.status]);
+					}
+
 					// Add to our tracking maps
 					this.tasks.set(task.id, task);
 					this.tasksBySession.set(sess.id, task.id);
@@ -526,6 +614,10 @@ export class BackgroundManager {
 
 	shutdown(): void {
 		this.shuttingDown = true;
+		if (this.refreshIntervalId) {
+			clearInterval(this.refreshIntervalId);
+			this.refreshIntervalId = undefined;
+		}
 		this.concurrency.clear();
 		this.notifications.clear();
 		try {
@@ -552,7 +644,7 @@ export class BackgroundManager {
 		} catch (error) {
 			if (task.status !== 'cancelled') {
 				task.status = 'error';
-				task.error = error instanceof Error ? error.message : 'Failed to acquire slot.';
+				task.error = extractErrorMessage(error, 'Failed to acquire slot.');
 				task.completedAt = new Date();
 				this.markForNotification(task);
 			}
@@ -579,6 +671,7 @@ export class BackgroundManager {
 					title: taskMetadata,
 				},
 				throwOnError: true,
+				...this.getClientOverrides(),
 			});
 			const session = unwrapResponse<{ id: string }>(sessionResult);
 			if (!session?.id) {
@@ -602,12 +695,26 @@ export class BackgroundManager {
 					parts: [{ type: 'text', text: task.prompt }],
 				},
 				throwOnError: true,
+				...this.getClientOverrides(),
 			});
 		} catch (error) {
-			this.failTask(
-				task,
-				error instanceof Error ? error.message : 'Failed to launch background task.'
-			);
+			const errorMsg = extractErrorMessage(error, 'Failed to launch background task.');
+			// Log the actual error for debugging — critical in sandbox environments
+			// where the client may silently fail due to missing baseUrl
+			try {
+				void this.ctx.client.app.log({
+					body: {
+						service: 'agentuity-coder',
+						level: 'error',
+						message: `Background task ${task.id} failed to start: ${errorMsg}`,
+					},
+					...this.getClientOverrides(),
+				});
+			} catch {
+				// If logging also fails, fall back to console
+				console.error(`[BackgroundManager] Task ${task.id} failed to start:`, errorMsg);
+			}
+			this.failTask(task, errorMsg);
 		}
 	}
 
@@ -686,29 +793,15 @@ export class BackgroundManager {
 
 	private async notifyParent(task: BackgroundTask): Promise<void> {
 		if (!task.parentSessionId) return;
+		if (this.shuttingDown) 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()}]
@@ -720,17 +813,36 @@ Task ID: ${task.id}
 
 Use the agentuity_background_output tool with task_id "${task.id}" to view the result.`;
 
-		try {
-			await this.ctx.client.session.prompt({
-				path: { id: task.parentSessionId },
-				body: {
-					parts: [{ type: 'text', text: message }],
-				},
-				throwOnError: true,
-				responseStyle: 'data',
-			});
-		} catch {
-			// Ignore notification errors
+		const maxRetries = 3;
+		for (let attempt = 0; attempt < maxRetries; attempt++) {
+			try {
+				await this.ctx.client.session.prompt({
+					path: { id: task.parentSessionId },
+					body: {
+						parts: [{ type: 'text', text: message }],
+					},
+					throwOnError: true,
+					responseStyle: 'data',
+					...this.getClientOverrides(),
+				});
+				// Mark as notified only AFTER confirmed delivery
+				notifiedStatuses.add(task.status);
+				task.notifiedStatuses = notifiedStatuses;
+				return; // Success
+			} catch (error) {
+				const errorMsg = extractErrorMessage(error, 'notification failed');
+				if (attempt < maxRetries - 1) {
+					// Exponential backoff: 1s, 2s, 4s
+					await new Promise((r) => setTimeout(r, 1000 * Math.pow(2, attempt)));
+					if (this.shuttingDown) return;
+				} else {
+					console.error(
+						`[BackgroundManager] Failed to notify parent for task ${task.id} after ${maxRetries} attempts:`,
+						errorMsg
+					);
+					// Don't mark as notified — allow future retry via refreshStatuses or Monitor
+				}
+			}
 		}
 	}
 
@@ -739,6 +851,7 @@ Use the agentuity_background_output tool with task_id "${task.id}" to view the r
 			await this.ctx.client.session.abort({
 				path: { id: sessionId },
 				throwOnError: false,
+				...this.getClientOverrides(),
 			});
 		} catch {
 			// Ignore abort errors
@@ -762,6 +875,7 @@ Use the agentuity_background_output tool with task_id "${task.id}" to view the r
 			const messagesResult = await this.ctx.client.session.messages({
 				path: { id: sessionId },
 				throwOnError: true,
+				...this.getClientOverrides(),
 			});
 			const messages = unwrapResponse<Array<unknown>>(messagesResult) ?? [];
 			const entries = Array.isArray(messages) ? messages : [];
@@ -910,3 +1024,24 @@ function unwrapResponse<T>(result: unknown): T | undefined {
 	}
 	return result as T;
 }
+
+/**
+ * Extract an error message from an unknown thrown value.
+ *
+ * The OpenCode SDK client (with `throwOnError: true`) throws **plain objects**
+ * (e.g. `{ message: "Not Found" }`) or raw strings rather than `Error` instances.
+ * This helper normalises all shapes into a usable string.
+ */
+function extractErrorMessage(error: unknown, fallback: string): string {
+	if (error instanceof Error) return error.message;
+	if (typeof error === 'string') return error || fallback;
+	if (typeof error === 'object' && error !== null) {
+		const obj = error as Record<string, unknown>;
+		if (typeof obj.message === 'string') return obj.message || fallback;
+		if (typeof obj.error === 'string') return obj.error || fallback;
+		if (typeof obj.error === 'object' && obj.error !== null) {
+			return extractErrorMessage(obj.error, fallback);
+		}
+	}
+	return fallback;
+}