[ { "name": "prompt quality penalties explain score delta", "input": { "Outcome": "completed", "OutcomeConfidence": "high", "Heuristics": { "ShortPromptCount": 1, "UnstructuredStart": true, "MissingSuccessCriteriaCount": 1, "MissingVerificationCount": 1, "NoCodeContextCount": 1 } }, "want_baseline_score": 100, "want_score": 94, "want_delta": -6, "want_grade": "A", "want_basis": [ "outcome", "prompt_quality", "context_quality" ], "want_penalties": { "constraintless_first_prompt": 1, "missing_success_criteria": 1, "code_task_without_context": 4 } }, { "name": "repeated prompts cap low confidence score delta", "input": { "Outcome": "unknown", "OutcomeConfidence": "low", "Heuristics": { "DuplicatePromptCount": 8 } }, "want_baseline_score": null, "want_score": 100, "want_delta": null, "want_grade": "A", "want_basis": [ "outcome", "prompt_quality" ], "want_penalties": null }, { "name": "workflow loop penalty adds workflow basis", "input": { "Outcome": "completed", "OutcomeConfidence": "high", "HasToolCalls": true, "RetryCount": 2, "Heuristics": { "RunawayToolLoopCount": 4 } }, "want_baseline_score": 90, "want_score": 85, "want_delta": -5, "want_grade": "B", "want_basis": [ "outcome", "tool_health", "workflow_quality" ], "want_penalties": { "tool_retries": 10, "repeated_failing_tool_cycles": 5 } } ]