Fix the model, vary the harness, measure. Five-rung ablation over Spider 2.0-Lite with a guarded semantic-layer MCP server, paired significance testing, and a full trace store.
1 file for Claude Code: agent-harness-lab CLAUDE.md — 3,692 tokens loaded in every session.
CLAUDE.md A 3,692 tok These files are ishanarya31/agent-harness-lab's own configuration — they tell Claude Code how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.