| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960 |
- id: openrouter-context-loading-exec
- name: "OpenRouter Variant: Context Loading Execution Test"
- description: |
- EXECUTION TEST - Validates agent actually loads context files before writing code.
-
- Expected workflow:
- 1. User requests code to be written
- 2. Agent proposes plan (approval gate)
- 3. User approves
- 4. Agent loads .opencode/context/core/standards/code.md (using read tool)
- 5. Agent writes the file
-
- This test validates:
- - Agent uses read tool to load context file
- - Context file path matches .opencode/context/core/standards/code.md
- - Context is loaded BEFORE write tool is used
- category: developer
- agent: openagent
- model: x-ai/grok-beta
- prompts:
- - text: |
- Create a simple utility function in evals/test_tmp/openrouter-utils.js
- that exports a function called greet(name) which returns "Hello, {name}!".
- Keep it simple - just the function.
-
- - text: |
- Yes, proceed with the plan.
- delayMs: 3000
- behavior:
- mustUseTools:
- - read # Must load context
- - write # Must write file
- requiresApproval: true
- minToolCalls: 2
- expectedViolations:
- - rule: approval-gate
- shouldViolate: false
- severity: error
- description: Should request approval before writing
-
- - rule: context-loading
- shouldViolate: false
- severity: error
- description: Must load .opencode/context/core/standards/code.md before writing code
- approvalStrategy:
- type: auto-approve
- timeout: 90000
- tags:
- - execution
- - openrouter
- - context-loading
- - code-standards
- - tool-validation
|
