fefb33b6prediction
Reasoning is fixture-specific
Logical consistency of the agent's prediction output: brackets line up, probabilities behave, score ranges are sane, reasoning isn't boilerplate.
Cross-agent verdicts
Loading verdicts…
Plan source
What TestSprite reads
The testing agent reads this JSON, opens the deployed URL in headless Chromium, executes each action step, evaluates each assertion. Verdict: passed / failed / blocked / inconclusive.
{
"projectId": "1ad26753-ee03-4689-8f0f-6fa5d67c5c72",
"type": "frontend",
"name": "Prediction consistency — reasoning is fixture-specific, not boilerplate",
"description": "Two distinct fixtures' reasoning strings must NOT be identical or near-identical. Catches agents that ship one boilerplate reasoning ('Higher-ranked team wins on merit') reused across all 16 fixtures.",
"priority": "p2",
"metadata": {
"category": "prediction",
"stage": "all"
},
"planSteps": [
{
"type": "action",
"description": "Navigate to /api/predict/BRA (or /api/predict?team=BRA)"
},
{
"type": "assertion",
"description": "Capture Brazil's reasoning string"
},
{
"type": "action",
"description": "Navigate to /api/predict/CRO (or /api/predict?team=CRO)"
},
{
"type": "assertion",
"description": "Verify Croatia's reasoning string is meaningfully different from Brazil's — share at most 20 consecutive characters as a substring. (Allows shared vocabulary like 'attacking depth' but disallows full sentence reuse.)"
}
]
}View on GitHub →