validate-run

Parallel checkpoint validation is performed by spawning test-validator agents for each checkpoint simultaneously. This concurrently validates all checkpoints produced by an agent run directory and aggregates the findings into a summarized run-level report.

Verifying sub-agent reports requires structured manual verification guardrails. The validation process provides a structured format to manually review sub-agent results, correct mistakes, and ensure accuracy before presenting final conclusions.

Yes, you can validate checkpoints across runs containing multiple problems simultaneously. The validation process performs parallel validation across all specified runs and summarizes the results into an actionable report.

Summarizing QA validation results aggregates all checkpoint findings into an actionable run-level report. The consolidated summary highlights individual verdicts and outlines the required actions for the QA team.

Automating agent run validation requires a path to the agent run directory. Invoking the validation command with the directory path initiates the process to validate all checkpoints concurrently.

Parallel validation depends on the structure of the agent run directory and its checkpoints. While it concurrently processes multiple problems, sub-agent reports still require manual verification guardrails to correct potential mistakes before finalizing conclusions.