metric-design

name	metric-design
description	Phase 1 of the auto-improvement loop — analyze the target repository, then design and calibrate the ruler (the trustworthy metric) BEFORE any optimization. Produces a low-variance primary metric, frozen anchors, a calibrated noise band, guardrails, reward-hack guards, and a mandatory canary that must clear the band or the ruler is rejected.
always	false
triggers	design the ruler, calibrate metric, metric design, noise band, canary check

metric-design — build the ruler (Phase 1)

Build the ruler before you measure anything with it. This skill drives Phase 1: analyze the target repository, then design a metric the keep-or-revert loop can trust. A loop optimizing a noisy or wrong ruler "wins" on fiction, and that is the dominant risk this app is built to eliminate — so Phase 2 does not start until Phase 1 has proven itself.

The output — the ruler, as durable config plus a reviewable doc

Writes data/ruler/: the calibrated ruler config and a human-readable metric-design document to review before any optimization runs. The ruler the active target profile supplies:

• Primary metric — a low-variance number where the two measurement arms cancel as much shared cost as possible (label + unit + direction). Never hard-coded in the UI; it is read from the profile.
• Attributable sub-stages — so a win is pinned to a named stage rather than hand-waved as a whole-system improvement.
• Frozen anchors — reference measurements with provenance (for example a pinned floor plus a shipped-defaults baseline), recorded in data/ruler/.
• Guardrails — metrics that must not regress beyond a stated tolerance.
• Reward-hack guards — checks the build/test gate structurally cannot see, such as "no silent capability shrink" or "a held-out functional probe still passes".

Calibration — the trust gate

1. Noise band — around 30 repetitions of the untouched baseline under the full harness, then noise_band = max(2σ, floor). Any delta inside the band is no change, not a small win.
2. Canary (mandatory) — a known or deliberately forced win that MUST clear the band. If it cannot, the harness is broken and the run halts. This is the Phase-1 gate: no Phase-2 cycle is trusted until the canary passes. The UI disables Start until the ruler reports status: "calibrated", and the backend independently refuses to run on an uncalibrated ruler — two checks because a UI-only guard is bypassable.

Why a rejected ruler is a good outcome

Halting on a failed canary feels like a failure and is the opposite. It means the measurement system caught its own untrustworthiness before spending a night optimizing noise. Report it plainly and say what would make the harness measurable instead.