[plan][feat]: SWE-bench Pro Evaluation Harness Integration #973
Description
Description
Placeholder for multi-agent planning in progress. This will be updated with the consensus plan.
Feature: Build a SWE-bench Pro evaluation harness that tests agentize lol impl pipeline against real-world software engineering tasks, including task ingestion, automated repository setup, isolated worktree execution, patch scoring, and metrics collection (tokens, accuracy, wall time).
Proposed Solution
Planning in progress via ultra-planner...
Related PR
TBD - will be updated when PR is created