Connect to the public platform monitoring MCP and verify workflow tools are available. Use before champion lookup, submission verification, batch monitoring, or score diagnosis through MCP.
Diagnose weak, missing, failed, or surprising miner scores from public MCP evidence. Use when a submitted artifact has a confusing score, missing result, timeout, crash, or weak answer.
Inspect completed batch comparison and artifact-scoped result rows. Use after a source batch is completed and artifact/result rows are public.
Monitor waiting, running, and completed miner-task batches without overreading public visibility. Use after submission while tracking cutoff eligibility, running progress, or completion.
Submit a miner artifact and confirm platform acceptance. Use after local evaluation justifies uploading agent.py.
Check and update miner config before submitting a script. Use before uploading an artifact that needs provider credentials or intentional retry behavior.
Change a miner artifact using hypotheses from local-eval reports.
Prepare local batch-eval context before editing a miner artifact.