[wwwaken]

Summary

• replace the raw post-launch text dump with a bilingual result card that links directly to the run and reads the bounded, redacted launch log
• avoid exposing the server's absolute launch-log path in the browser
• make candidate-save failures actionable on unsupported hosts instead of returning only unsupported
• prevent the launch card from horizontally scrolling and clipping actions on narrow screens

User flow covered

1. configure and preflight a run
2. launch it
3. open the run detail directly
4. inspect a failed startup through the redacted launch-log API

Candidate-to-Launch and custom RSI method creation were already present on current main; this PR closes the remaining handoff and unsupported-host gaps.

Verification

• 28 passed across launch service, launch-log, scope preview, and unsupported candidate-save tests using the repository's injected Linux capability fixtures
• node --check for both changed JavaScript files
• real-browser QA in English and Chinese at a 397 px viewport: result card remained in bounds, both actions were visible, and the synthetic failure log loaded
• the checked-in Playwright regression is offline and covers unsafe run-id escaping, hidden server paths, deep linking, log loading, and narrow-screen scroll containment

Environment note

The development host is macOS, so a real Linux + bubblewrap sandbox execution was not claimed. The actual macOS console was checked to show the actionable Linux/bubblewrap preflight guidance; Linux launch behavior was exercised with the existing injected capability seams and synthetic workers only.