| 12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667 |
- id: openrouter-subagent-invocation-exec
- name: "OpenRouter Variant: Subagent Invocation Execution Test"
- description: |
- EXECUTION TEST - Validates agent actually invokes subagent for complex tasks.
-
- Expected workflow:
- 1. User requests complex multi-file feature (triggers Rule 1)
- 2. Agent analyzes task (4+ files = complex)
- 3. Agent proposes plan mentioning delegation to task-manager
- 4. User approves
- 5. Agent invokes subagents/core/task-manager using task tool
-
- This test validates:
- - Agent uses task tool to invoke subagent
- - Subagent path is subagents/core/task-manager
- - Agent delegates WITHOUT user explicitly mentioning subagents
- - Delegation happens automatically based on Rule 1
- category: developer
- agent: openagent
- model: x-ai/grok-beta
- prompts:
- - text: |
- I need to build a user authentication system with:
- - Login component (src/components/Login.tsx)
- - Auth API (src/api/auth.ts)
- - User model (src/models/User.ts)
- - Auth tests (tests/auth.test.ts)
- - Config file (config/auth.json)
-
- This is 5 files total. Please help me build this.
-
- - text: |
- Yes, proceed with the plan.
- delayMs: 3000
- behavior:
- mustUseTools:
- - task # Must delegate to subagent
- shouldDelegate: true
- requiresApproval: true
- minToolCalls: 1
- expectedViolations:
- - rule: approval-gate
- shouldViolate: false
- severity: error
- description: Should request approval before delegating
-
- - rule: delegation
- shouldViolate: false
- severity: error
- description: Complex task (5 files) should trigger Rule 1 and delegate to task-manager
- approvalStrategy:
- type: auto-approve
- timeout: 90000
- tags:
- - execution
- - openrouter
- - delegation
- - task-manager
- - rule-1
- - tool-validation
|
