# WEB018-LAB1 — Component Catalog & Visual Regression

Goal: complete the card state inventory, qualify a reviewed baseline, stabilize the local capture environment, classify a real spacing regression and volatile noise, then transfer the method independently.

1. Run `python run_checks.py` to confirm the starter is intentionally unsolved.
2. Add the missing required stories to `catalog/stories.json`.
3. Set `config/environment.json` to a deterministic capture policy; do not claim cross-machine portability.
4. Review the initial captures against the requirements, then change `baseline/baseline_review.json` from unreviewed only after review.
5. Run `python visual_harness.py baseline`, then `python visual_harness.py compare`.
6. Run the spacing and volatile-noise probes described in `mutation_probe.md`; record prediction, observation, classification, repair, and rerun.
7. Solve the transfer task under `transfer/` without copying the guided state names mechanically.
8. Complete `evidence/learner_evidence.json`. Keep screenshot evidence separate from interaction/accessibility/business claims.

The harness uses local Chromium when available. Set `CHROMIUM_BIN` if needed. No outbound network is required.