Problem
The automation health dashboard reports on link-health and dep-bump impact, but has no visibility into the clawgenti PR reviewer's effect on the org. We cannot currently answer "is the PR reviewer helping PRs merge faster, and how much is it reviewing?"
The PR-review bot already emits queue, history, and per-run throughput data, but it does not compute the one signal needed for impact: time-to-merge segmented by whether a PR was reviewed by the bot, and segmented before vs. after the bot began covering each repo.
Rationale
Repos were onboarded to the reviewer at different times, so a single global "before/after" date would be misleading. The impact measure must be per-repo, keyed to when each repo first received a bot review (a queryable signal), then aggregated.
Scope (this repo)
Produce a small, read-only metrics step that queries merged PRs across covered repos, derives each repo's activation from its earliest reviewed PR, and writes an impact.json report (reviewed vs. unreviewed and before vs. after-activation median time-to-merge). The health dashboard (tracked separately in kagenti/agent-skills) consumes this report.
Acceptance
- New report captures reviewed/unreviewed and before/after-activation median TTM with per-repo activation dates
- No hardcoded split date; activation derived at runtime
- Unit-tested segmentation/aggregation math
Problem
The automation health dashboard reports on link-health and dep-bump impact, but has no visibility into the clawgenti PR reviewer's effect on the org. We cannot currently answer "is the PR reviewer helping PRs merge faster, and how much is it reviewing?"
The PR-review bot already emits queue, history, and per-run throughput data, but it does not compute the one signal needed for impact: time-to-merge segmented by whether a PR was reviewed by the bot, and segmented before vs. after the bot began covering each repo.
Rationale
Repos were onboarded to the reviewer at different times, so a single global "before/after" date would be misleading. The impact measure must be per-repo, keyed to when each repo first received a bot review (a queryable signal), then aggregated.
Scope (this repo)
Produce a small, read-only metrics step that queries merged PRs across covered repos, derives each repo's activation from its earliest reviewed PR, and writes an
impact.jsonreport (reviewed vs. unreviewed and before vs. after-activation median time-to-merge). The health dashboard (tracked separately in kagenti/agent-skills) consumes this report.Acceptance