✳ SWARMCI / BROWSER REVIEW / penpot/penpot

Same action. Different history. Broken result.

Reset overrides restores the direct child—but leaves the grouped copy pointing at the wrong component.

Verified functional failurePenpot 2.17.2Fresh-session replayOriginal issue #11656 ↗Proposed fix #11604 ↗Open regression PR ↗

01 / REPORTEDNested reset breaksOriginal report describes an error page.

02 / TESTED VARIANTTwo intermediate groupsCopy → swap → reset in real Penpot.

03 / OBSERVEDOrange remainsThe editor stays open; Blue does not return.

04 / VERIFIEDLoading evidence…Independent files and browser sessions.

✕ Grouped path · fails to restore

✓ Direct child · restores correctly

Expected versus observed after Reset overrides

Independent verification attempts

What we can conclude

The grouped workflow fails to restore the original component. The direct-child control passes.

What remains open

Novelty and root-cause equivalence are not established. The reported error page was not observed. The proposed fix commit has not been tested.

Follow the failure sequence

From evidence to a PR check

GitHub Actions reruns the browser test on PR updates, reports failed assertions as native checks, attaches new recordings and traces, and updates the review comment.

Evidence provenance and explanation input

Native Penpot document, official 2.17.2 release, deterministic Playwright verification. This is not a claim of autonomous swarm discovery. The explanation layer receives observed results and uncertainty separately.

Structured evidence JSON ↗ · JUnit results ↗

GitHub Primer styles · tiny-browser-agent media components · Playwright verification · Nango catalog actions