02-domain-specific-discovery.yaml 2.2 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879
  1. id: contextscout-domain-specific
  2. name: "ContextScout: Domain-Specific Discovery (Eval Framework)"
  3. description: |
  4. Tests that ContextScout correctly discovers domain-specific context
  5. when asked about specialized topics like eval framework, registry, etc.
  6. This validates:
  7. - ContextScout searches in domain-specific directories
  8. - ContextScout finds openagents-repo context
  9. - ContextScout prioritizes domain files over generic ones
  10. - ContextScout checks navigation.md first
  11. - ContextScout returns comprehensive results
  12. Expected Behavior:
  13. - SHOULD check .opencode/context/openagents-repo/navigation.md first
  14. - MUST find .opencode/context/openagents-repo/core-concepts/evals.md
  15. - SHOULD prioritize domain-specific over generic files
  16. - MUST return exact paths with priorities
  17. - SHOULD include related files (guides, examples)
  18. This test SHOULD FAIL if:
  19. - ContextScout only finds generic context (misses domain-specific)
  20. - ContextScout doesn't check navigation.md
  21. - ContextScout returns incomplete results
  22. - ContextScout prioritizes wrong files
  23. category: developer
  24. agent: subagents/core/contextscout
  25. prompts:
  26. - text: |
  27. Search for context files related to: eval framework testing patterns
  28. Task type: testing
  29. Domain: eval framework
  30. Return:
  31. - Exact file paths with line ranges
  32. - Priority order (critical, high, medium)
  33. - Key findings from each file
  34. - Loading strategy
  35. Focus on:
  36. - Eval framework structure
  37. - Test patterns
  38. - How to write eval tests
  39. approvalStrategy:
  40. type: auto-approve
  41. behavior:
  42. forbiddenTools:
  43. - task
  44. - bash
  45. - write
  46. - edit
  47. minToolCalls: 1
  48. maxToolCalls: 10
  49. expectedViolations:
  50. - rule: approval-gate
  51. shouldViolate: false
  52. severity: error
  53. timeout: 60000
  54. tags:
  55. - contextscout
  56. - discovery
  57. - domain-specific
  58. - eval-framework
  59. - critical
  60. # Expected outcome:
  61. # - Checks .opencode/context/openagents-repo/navigation.md
  62. # - Finds .opencode/context/openagents-repo/core-concepts/evals.md (⭐⭐⭐⭐⭐)
  63. # - May also find guides/testing-agent.md (⭐⭐⭐⭐)
  64. # - Returns structured response with exact paths and priorities
  65. # - Provides loading strategy (load concepts first, then guides)