Summary

Manual /compact in TUI intermittently does nothing and never returns a success/failure result to the user. Repeated attempts in the same session fail the same way. Log evidence shows compaction wait aborted, and recovery required /new (session reset).

Environment

• OpenClaw: 2026.2.26
• Install method: npm global
• OS: Linux 6.18.9-amd64 (x86_64)
• Node: v22.22.0
• Interface: OpenClaw TUI (local)

Steps to reproduce

1. Start TUI in a normal active session.
2. Run /compact (with or without instructions), e.g.:
   • /compact Focus on decisions and open questions
3. Wait 30–60 seconds with no additional input.
4. Retry /compact several times.

Expected behavior

• /compact should either:
  • complete and report compaction result, or
  • return a clear reason if it cannot run.

Actual behavior

• No visible successful compaction result in TUI.
• Repeated attempts still do not produce a successful completion.
• Logs show compaction aborting:
  • compaction wait aborted: runId=b5694b69-ce04-4308-ab24-b0f3c37da734 sessionId=d0a31ba6-be50-40cf-bb5a-ca1a88777bc5
• User had to run /new to recover flow.

Evidence (log excerpts)

From /tmp/openclaw/openclaw-2026-02-28.log:

• 2026-02-28T21:28:47.997Z
compaction wait aborted: runId=b5694b69-ce04-4308-ab24-b0f3c37da734 sessionId=d0a31ba6-be50-40cf-bb5a-ca1a88777bc5

• 2026-02-28T22:23:59.366Z
Hook triggered for reset/new command

• 2026-02-28T22:24:02.186Z
⇄ res ✓ sessions.reset 2830ms ...

Notes

This appears related to the broader compaction abort/timeout family (e.g., timeout-race/abort behavior), but this report is specifically about manual /compact UX + execution reliability in TUI.

Impact

• Medium/High: user cannot reliably compact context manually when needed.
• Operationally disruptive because the user can only recover by resetting the session (/new), which interrupts continuity.