You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
fix: detect crashed agent immediately instead of waiting 10min timeout
Previously, if the agent process died before writing any log entries
(e.g. startup crash), the dead-process check was skipped because it
required entries.length > 0. This left sessions stuck for the full
10-minute safety valve timeout. Also, agent stderr was piped to
/dev/null, making crash debugging impossible.
- Capture agent stderr to agent-stderr.log instead of /dev/null
- Remove entries.length > 0 guard from dead-process detection
- Read stderr in both crash detection and safety valve paths for logging
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
0 commit comments