Summary / Harness arms
A5 h2p (repair prompt)
The gate bundle with the `repair_v1` prompt: reproduce first, read before editing, make the smallest change.
Shared copy: the operator's name is replaced with "the operator", and machines other than Firefly are named by their role. Nothing else is changed.
What is built now
Registered (prompt written, arm config stored).
What is missing
Never run.
Reviewer Claude, 2026-10-04
Reviewer's read: Registered, never run
Deliberately behind the vote in the queue. With dev saturated it now needs the harder set.
How it is compared not run
Against h1-v41 on the harder set.
Adoption rules shared by every arm: How the tests work.
Research & attempts
No documents yet.
Score history
No scores yet. The first entry is the baseline for suite v1.
Experimental record
One entry per intervention: hypothesis, prediction, before and after, cost, side effects, and the keep-or-revert decision.
No interventions recorded yet.