Review depth: normalise comment count by PR size (lines changed)

[gkanitz]

Parent

#1

Background

The report currently classifies reviews as shallow (0–2 comments) or deep (≥3 comments) using an absolute comment count threshold. This treats a 3-comment review on a 10-line fix the same as a 3-comment review on a 500-line refactor, which understates the engagement on large PRs and overstates it on trivial ones.

What to build

Normalise the shallow/deep split by the size of the reviewed PR: compute a comments-per-100-lines-changed ratio and threshold against that instead of (or in addition to) the absolute count.

Why deferred

Requires storing `additions` + `deletions` from the PR object for every PR the subject reviewed — a schema change to the activity set and an adapter change to fetch the extra field. GitHub exposes this without a diff fetch (it is on the PR object itself), but the scope increase (schema, adapter, tests, renderer) was judged out of proportion for v1. The ≥3 absolute threshold is a good-enough proxy for v1.

Acceptance criteria

• PR size (additions + deletions) is fetched and stored per reviewed PR (GitHub and GitLab adapters)
• Shallow/deep threshold is expressed as comments per N lines changed (configurable or well-chosen constant), not absolute count
• The headline "deep review %" KPI and the stacked timeline chart both use the normalised threshold
• Existing ≥3 absolute threshold tests are updated or replaced

Blocked by

• Visualisation / charting issue (graphs overhaul, to be filed)

[gkanitz]

Scheduled: implemented by #52 in the v0.2 "Meaning" wave (spec: docs/superpowers/specs/2026-07-04-impact-roadmap-design.md, PR #47). The normalization formula is pinned there: deep review ⇔ comments ≥ clamp(ceil(lines/100), 3, 10), with an honest fallback basis for PRs without diff data. The PR closing #52 will close this issue.