← ClaudeAtlas

baseline-reviewlisted

Review E2E baseline image changes in the current PR.
gpambrozio/Gallager · ★ 3 · Code & Development · score 69
Install: claude install-skill gpambrozio/Gallager
# E2E Baseline Image Review Analyze all E2E baseline image changes in the current PR, compare them against `main`, and produce a report indicating whether each change is expected given the PR description. ## Overview E2E baselines live in `E2ETests/<scenario-name>/##-<label>.png`. PRs that change UI behavior regenerate these screenshots. A typical PR with UI changes can touch 200+ baseline images, but most are just dithering/anti-aliasing artifacts from re-rendering. This skill separates the noise from real changes and verifies that real changes match the PR intent. ## Known Patterns Before reviewing, be aware of these recurring patterns in baseline changes: **Random pairing codes:** Images named `*-mac-code-generated.png` show the macOS pairing code screen. The 6-letter code is randomly generated each test run, so these images always differ between runs. This is expected and normal — not a rendering artifact, just non-deterministic test content. Classify these as "expected (non-deterministic content)." **iOS paired/connected screens:** Images named `*-ios-paired.png` and `*-mac-connected.png` often have sub-pixel dithering differences. The script usually catches these as dithering, but if they show up as "changed" with very low diff percentages (<0.5%), they're almost certainly just rendering noise. **Platform prefixes:** Baseline filenames always start with a platform prefix after the auto-numbered counter — `mac-` / `ios-` for standard scenarios, and `host-` / `vie