## Summary
- Replaced `deflake-e2e-recent-prs` command with
`deflake-e2e-recent-commits` that scans CI workflow runs on main instead
of PR comments
- Downloads the `html-report` artifact (`results.json`) from each CI run
to extract flaky test data, which works for push events that don't post
PR comments
- Updated `claude-deflake-e2e.yml` workflow to use the new command

## Test plan
- [ ] Trigger the `Claude Deflake E2E` workflow manually and verify it
correctly scans recent main branch CI runs
- [ ] Verify flaky tests are correctly parsed from `results.json`
artifacts

🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- devin-review-badge-begin -->

---

<a href="https://app.devin.ai/review/dyad-sh/dyad/pull/2607"
target="_blank">
  <picture>
<source media="(prefers-color-scheme: dark)"
srcset="https://static.devin.ai/assets/gh-open-in-devin-review-dark.svg?v=1">
<img
src="https://static.devin.ai/assets/gh-open-in-devin-review-light.svg?v=1"
alt="Open with Devin">
  </picture>
</a>
<!-- devin-review-badge-end -->

<!-- CURSOR_SUMMARY -->
---

> [!NOTE]
> **Low Risk**
> Low risk doc/workflow tweak that changes how the deflake automation
sources flaky tests (GitHub Actions runs/artifacts) but does not touch
production code or test logic.
> 
> **Overview**
> Updates the deflake automation to **scan recent `main` CI workflow
runs** (push events) instead of PR Playwright summary comments, by
downloading each run’s `html-report` artifact and parsing `results.json`
to detect retry-passed tests with prior failures/timeouts.
> 
> Adjusts the scheduled `Claude Deflake E2E` workflow to accept
`commit_count`, grant `actions: read`, and invoke
`/dyad:deflake-e2e-recent-commits` rather than the old PR-based command.
> 
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
0da1e67da43e509577d5b8dc1f155779742d1529. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->

<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Switched the deflake command to scan recent main CI runs and parse
html-report results.json to find flaky E2E tests. Updated the Claude
Deflake E2E workflow to use commit_count and added actions: read
permission.

- **Refactors**
- List completed main push runs via gh api, fetch 3x commit_count, and
filter to success/failure.
- Download non-expired html-report artifacts; parse results.json with a
Node.js script to detect flakes (final passed after
fail/timedOut/interrupted).
- Build "<spec_path.spec.ts> > Suite > Test" titles; group and rank by
frequency; clean up artifacts.
  - Skip runs without artifacts; note 3-day artifact retention.

- **Bug Fixes**
  - Updated command doc to reference the TodoWrite tool.

<sup>Written for commit 0da1e67da43e509577d5b8dc1f155779742d1529.
Summary will update on new commits.</sup>

<!-- End of auto-generated description by cubic. -->

---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>