Add Browserbase workflow eval docs

[rforgeon]

Summary

Adds a small harness-neutral eval set for Browserbase browser automation workflows.

What changed

• Add eval cases for safe navigation, trace-to-API analysis, and UI regression testing
• Add a rubric covering workflow selection, boundary safety, evidence quality, and privacy boundaries
• Document optional Telvine publishing guidance with metadata-only telemetry events

Validation

for f in evals/*/cases.jsonl; do jq -e . "$f" >/dev/null || exit 1; done
rg -n "Evals and production telemetry|evals/" README.md
git diff --check

Updated validation wording after Cursor Bugbot flagged that line-by-line jq validation can fail on blank JSONL lines.

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 6cb7602. Configure here.}

[cursor]

Blank line breaks JSONL validation

Low Severity

cases.jsonl includes a fourth empty line after the three JSON records. The PR’s validation loop runs jq on every line read from the file, so that blank line makes jq fail and the documented check exits with an error even though the three cases are valid JSON.

 

^{Reviewed by Cursor Bugbot for commit 6cb7602. Configure here.}

[rforgeon]

Thanks Bugbot. The branch file has three JSONL records with no trailing blank record, but the original PR validation snippet was stricter than needed for JSONL. I updated the PR body to validate with jq -e . "$f", which handles normal JSONL whitespace safely.