Autoresearch: Phase 1 - Core harness deployed on autoresearch/pro/harness

[realityinspector]

The autoresearch harness is live on branch autoresearch/pro/harness.

What's built:

• pro_autoresearch.py — main loop: mutate config -> run simulation -> extract metrics -> keep/discard
• config_space.py — 27 mutable dimensions across 6 mechanism clusters
• metrics.py — Causal Resolution metric (Coverage x Convergence), dry-run synthetic metrics
• pareto.py — Pareto frontier analysis (quality vs cost)

Verified: Dry-run mode: 10 iterations in <1 second, produces valid JSONL + Pareto JSON

Next: Pro-1 through Pro-8 swarm agents running cluster-specific optimization.