refactor: separate sub-agent workflows, update model to Opus 4.6, remove hourly control loop

Addresses feedback from issue strands-agents#33 comment:

1. Remove hourly control loop (strands-control.yml)
   - Not needed for our use case. The Wednesday 10am cron is directly
     in strands-autonomous.yml — no AGENT_SCHEDULES machinery required.

2. Update model to Opus 4.6 with adaptive thinking and 1M context
   - Model: global.anthropic.claude-opus-4-6-v1
   - Max tokens: 128,000 (was 64,000)
   - Thinking: adaptive (was fixed 8k budget)
   - Removed STRANDS_BUDGET_TOKENS constant
   - Matches strands-coder settings

3. Separate sub-agent workflows for isolated testing
   - strands-adversarial-test.yml — adversarial testing agent
   - strands-release-notes-agent.yml — release notes generation
   - strands-docs-gap.yml — documentation gap analysis
   - Each can be triggered independently via workflow_dispatch
   - Orchestrator dispatches to dedicated workflows

4. Bug fix: _get_all_tools() had dead code after return statement
   - Orchestrator tools were never actually loaded
   - Fixed to return after appending all tools

Files: 6 modified, 3 new, 1 deleted

Author: agent-of-mkmeral

strands-command/README.md

@@ -544,20 +544,24 @@ The orchestrator module (`orchestrator.py`) enables agents to dispatch and coord
 
 ```mermaid
 graph TD
-    A[Control Loop] -->|hourly check| B[Schedule Check]
-    B -->|Wednesday 10am| C[Release Digest Agent]
-    C -->|dispatch| D[Adversarial Tester]
-    C -->|dispatch| E[Release Notes Generator]
-    C -->|dispatch| F[Docs Gap Analyzer]
+    A[strands-autonomous.yml] -->|"Wed 10am cron"| C[Release Digest Orchestrator]
+    C -->|dispatch| D[strands-adversarial-test.yml]
+    C -->|dispatch| E[strands-release-notes-agent.yml]
+    C -->|dispatch| F[strands-docs-gap.yml]
     D -->|results| C
     E -->|results| C
     F -->|results| C
     C -->|create| G[Digest Issue]
     
     style C fill:#f9f,stroke:#333,stroke-width:2px
     style G fill:#9f9,stroke:#333,stroke-width:2px
+    style D fill:#fcc,stroke:#333
+    style E fill:#ccf,stroke:#333
+    style F fill:#cfc,stroke:#333
 ```
 
+Each sub-agent workflow can also be triggered independently via `workflow_dispatch` for isolated testing.
+
 #### Self-Trigger Prevention
 
 The orchestrator uses the same pattern as strands-coder to prevent infinite loops:
@@ -575,26 +579,12 @@ Copy the template workflows to your repository:
 ```bash
 # From the devtools/strands-command/workflows/ directory
 cp strands-autonomous.yml your-repo/.github/workflows/
-cp strands-control.yml your-repo/.github/workflows/
+cp strands-adversarial-test.yml your-repo/.github/workflows/
+cp strands-release-notes-agent.yml your-repo/.github/workflows/
+cp strands-docs-gap.yml your-repo/.github/workflows/
 ```
 
-#### 2. Configure Repository Variables
-
-Set the `AGENT_SCHEDULES` repository variable:
-
-```json
-{
-  "jobs": {
-    "weekly_release_digest": {
-      "enabled": true,
-      "cron": "0 10 * * 3",
-      "command": "release-digest",
-      "workflow": "strands-autonomous.yml",
-      "last_triggered": 0
-    }
-  }
-}
-```
+The Wednesday 10am schedule is configured directly in `strands-autonomous.yml` via cron — no additional schedule variables needed.
 
 #### 3. Configure Secrets
 
@@ -649,3 +639,57 @@ When running as an orchestrator agent, these tools are available:
 | `check_agents_status` | Check status of all dispatched sub-agents |
 | `wait_for_agents` | Wait for all sub-agents to complete (with timeout) |
 | `get_orchestrator_config` | View current orchestrator security configuration |
+
+
+### Autonomous Agent Capabilities
+
+The Strands command system supports autonomous agent execution for:
+
+- **Adversarial Testing**: Automated testing that actively tries to break code changes
+- **Release Digest**: Weekly orchestrated analysis combining adversarial testing, release notes, and documentation gap analysis
+
+#### Workflow Architecture
+
+Each agent type has its own dedicated workflow for **isolated testing and execution**:
+
+| Workflow | Purpose | Trigger |
+|----------|---------|---------|
+| `strands-autonomous.yml` | Release digest orchestrator | Wednesday 10am UTC cron + manual |
+| `strands-adversarial-test.yml` | Adversarial testing agent | Dispatched by orchestrator + manual |
+| `strands-release-notes-agent.yml` | Release notes generation | Dispatched by orchestrator + manual |
+| `strands-docs-gap.yml` | Documentation gap analysis | Dispatched by orchestrator + manual |
+
+#### Agent Orchestration
+
+The orchestrator module (`orchestrator.py`) enables agents to dispatch and coordinate sub-agents via dedicated workflows:
+
+```
+strands-autonomous.yml (Release Digest Orchestrator — Wed 10am)
+  ├── strands-adversarial-test.yml      (isolated workflow)
+  ├── strands-release-notes-agent.yml   (isolated workflow)
+  └── strands-docs-gap.yml             (isolated workflow)
+```
+
+Each sub-agent runs in its own workflow, so you can:
+- **Test in isolation**: Trigger any sub-agent independently via `workflow_dispatch`
+- **Debug independently**: Check workflow run logs per agent type
+- **Customize timeouts**: Each workflow has its own `timeout-minutes`
+
+Security limits:
+- Max concurrent sub-agents: 3 (configurable)
+- Per-agent timeout: 30 minutes
+- Max total sub-agents per run: 5
+- Rate limiting and cooldown between dispatches
+
+#### Model Configuration
+
+All agents use **Opus 4.6** with adaptive thinking and 1M context window:
+- Model: `global.anthropic.claude-opus-4-6-v1`
+- Max tokens: 128,000
+- Thinking: adaptive
+
+#### Setup
+
+1. Copy the workflow files to `.github/workflows/` in your repository
+2. Configure `PAT_TOKEN` secret with `workflow_dispatch` permission
+3. Configure orchestrator limits via repository variables (optional)

strands-command/agent-sops/task-release-digest.sop.md

@@ -64,11 +64,11 @@ Determine which sub-agents to dispatch based on the changes found.
 **Constraints:**
 - You MUST plan the following sub-agent tasks:
 
-  | Task | Agent Type | Input | Output |
-  |------|-----------|-------|--------|
-  | Adversarial Testing | `adversarial-test` | List of PRs with significant code changes | Findings report per PR |
-  | Release Notes | `release-notes` | Base and head git references | Formatted release notes |
-  | Documentation Gaps | `docs-gap` | List of PRs with new/changed APIs | Missing docs report |
+  | Task | Agent Type | Workflow | Input | Output |
+  |------|-----------|----------|-------|--------|
+  | Adversarial Testing | `adversarial-test` | `strands-adversarial-test.yml` | List of PRs with significant code changes | Findings report per PR |
+  | Release Notes | `release-notes` | `strands-release-notes-agent.yml` | Base and head git references | Formatted release notes |
+  | Documentation Gaps | `docs-gap` | `strands-docs-gap.yml` | List of PRs with new/changed APIs | Missing docs report |
 
 - You MUST skip adversarial testing if there are no code-changing PRs (only docs/CI/test changes)
 - You MUST skip documentation gap analysis if there are no API-changing PRs
@@ -80,10 +80,14 @@ Determine which sub-agents to dispatch based on the changes found.
 Dispatch sub-agents for parallel execution.
 
 **Constraints:**
-- You MUST use the orchestrator module (`orchestrator.py`) to dispatch sub-agents
+- You MUST use the orchestrator module (`orchestrator.py`) to dispatch sub-agents to their **dedicated workflows** (each runs in isolation)
 - You MUST call the `dispatch_agent` function for each planned task
 - You MUST respect the concurrent agent limit — wait for slots before dispatching
 - You MUST wait the minimum cooldown between dispatches
+- You MUST dispatch each sub-agent to its dedicated workflow:
+  - **Adversarial tester** → `strands-adversarial-test.yml`
+  - **Release notes** → `strands-release-notes-agent.yml`
+  - **Docs gap** → `strands-docs-gap.yml`
 - You MUST pass appropriate inputs to each sub-agent:
   - **Adversarial tester**: PR numbers, branch references
   - **Release notes**: Base tag, head reference, repository
@@ -228,7 +232,7 @@ The orchestrator reads scheduling configuration from the `AGENT_SCHEDULES` repos
       "cron": "0 10 * * 3",
       "prompt": "Run the weekly release digest. Find all changes since the last release, dispatch adversarial testing and release notes sub-agents, and compile a comprehensive digest issue.",
       "system_prompt": "You are a Release Digest Orchestrator following the task-release-digest SOP.",
-      "workflow": "strands-command.yml",
+      "workflow": "strands-autonomous.yml"  // orchestrator only,
       "tools": ""
     }
   }

strands-command/scripts/python/agent_runner.py

@@ -52,10 +52,10 @@
 from notebook import notebook
 from str_replace_based_edit_tool import str_replace_based_edit_tool
 
-# Strands configuration constants
-STRANDS_MODEL_ID = "global.anthropic.claude-opus-4-5-20251101-v1:0"
-STRANDS_MAX_TOKENS = 64000
-STRANDS_BUDGET_TOKENS = 8000
+# Strands configuration constants — matches strands-coder settings
+# Opus 4.6 with adaptive thinking and 1M context window
+STRANDS_MODEL_ID = "global.anthropic.claude-opus-4-6-v1"
+STRANDS_MAX_TOKENS = 128000
 STRANDS_REGION = "us-west-2"
 
 # Default values for environment variables used only in this file
@@ -157,7 +157,7 @@ def _get_trace_attributes() -> dict:
     }
 
 def _get_all_tools() -> list[Any]:
-    return [
+    tools = [
         # File editing
         str_replace_based_edit_tool,
 
@@ -191,6 +191,8 @@ def _get_all_tools() -> list[Any]:
     # Add orchestrator tools if available
     if _ORCHESTRATOR_AVAILABLE:
         tools.extend([dispatch_agent, check_agents_status, wait_for_agents, get_orchestrator_config])
+    
+    return tools
 
 
 def run_agent(query: str):
@@ -203,13 +205,10 @@ def run_agent(query: str):
         # Get tools and create model
         tools = _get_all_tools()
 
-        # Create Bedrock model with inlined configuration
+        # Create Bedrock model — Opus 4.6 with adaptive thinking and 1M context
         additional_request_fields = {}
-        additional_request_fields["anthropic_beta"] = ["interleaved-thinking-2025-05-14"]
-        
         additional_request_fields["thinking"] = {
-            "type": "enabled",
-            "budget_tokens": STRANDS_BUDGET_TOKENS
+            "type": "adaptive",
         }
 
         model = BedrockModel(

strands-command/scripts/python/orchestrator.py

@@ -64,7 +64,7 @@ class AgentTask:
     completion_time: datetime | None = None
     result: str | None = None
     error: str | None = None
-    workflow: str = "strands-command.yml"
+    workflow: str = ""
 
 
 @dataclass
@@ -170,7 +170,7 @@ def dispatch_agent(
         agent_type: str,
         prompt: str,
         system_prompt: str = "",
-        workflow: str = "strands-command.yml",
+        workflow: str = "",
         extra_inputs: dict[str, str] | None = None,
     ) -> AgentTask:
         """Dispatch a sub-agent via workflow_dispatch.
@@ -179,7 +179,10 @@ def dispatch_agent(
             agent_type: Type of agent (e.g., "adversarial-test", "release-notes")
             prompt: Task prompt for the sub-agent
             system_prompt: System prompt override for the sub-agent
-            workflow: Target workflow file (default: strands-command.yml)
+            workflow: Target workflow file. Each sub-agent has a dedicated workflow:
+                  - strands-adversarial-test.yml
+                  - strands-release-notes-agent.yml
+                  - strands-docs-gap.yml
             extra_inputs: Additional workflow inputs
 
         Returns:
@@ -468,7 +471,7 @@ def dispatch_agent(
     agent_type: str,
     prompt: str,
     system_prompt: str = "",
-    workflow: str = "strands-command.yml",
+    workflow: str = "",
 ) -> str:
     """Dispatch a sub-agent via GitHub Actions workflow_dispatch.
 
@@ -482,7 +485,10 @@ def dispatch_agent(
         agent_type: Type of agent (e.g., "adversarial-test", "release-notes", "docs-gap")
         prompt: Task prompt for the sub-agent
         system_prompt: System prompt override (optional)
-        workflow: Target workflow file (default: strands-command.yml).
+        workflow: Target workflow file. Each sub-agent has a dedicated workflow:
+                  - strands-adversarial-test.yml
+                  - strands-release-notes-agent.yml
+                  - strands-docs-gap.yml.
                   Use "owner/repo/workflow.yml" for cross-repo dispatch.
 
     Returns: