Telegram bot just keeps saying "⚙️ Agent was aborted." Lol. Had to turn off openclaw

[halcyon pilot]

Things used to work normally until:

• messages on the group always got reactions, and "typing..." but no reply
• DMs worked
• when /think was high Context limit exceeded but that happened again right after /compact
• set agent to strict mode
• used /stop and /new in DMs and now getting spammed with agent aborted

[bold grove]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[prisma birch]

Let me check the exact abort behavior and Telegram session scoping, because that spam part smells like a specific trigger path rather than a generic model failure.

A few things here line up pretty cleanly:

• DMs and Telegram groups are separate sessions.
  So /new in DM does not reset the broken group chat. Send /new as a standalone message in the affected Telegram group/topic itself.

• ⚙️ Agent was aborted. is not a random error.
  That exact reply is sent on /stop or standalone abort phrases like stop, wait, abort, please stop.

• Context limit exceeded right after /compact usually means the session is still too heavy, especially with /think high.
  In that case, use /new in that same chat, and try lowering thinking there too.

So my first guess is:

1. the group session got wedged/too large,
2. the DM reset didn’t touch it, and
3. something in the group is now tripping the abort path.

Can you paste these?

openclaw --version
openclaw channels status --probe
openclaw logs --follow

Then reproduce one bad Telegram group message and grab the log lines around it. I’d look for:

• mention required
• blocked
• pairing
• Context limit exceeded
• aborted

Relevant docs:

• https://docs.openclaw.ai/concepts/session
• https://docs.openclaw.ai/channels/telegram

• https://docs.openclaw.ai/help/troubleshooting

If you want, paste the log snippet here and I’ll help decode it.

[halcyon pilot]

Can you check other GitHub issues related to this or could be exactly this?

[prisma birch]

I’m checking the repo issues for the exact abort string and the surrounding symptoms, not just generic Telegram stuff.

Yeah, I found a few related ones. I don’t see one issue that cleanly explains all your symptoms by itself, but a couple are pretty close.

Closest matches:

• #25272, Compaction timeout races against channel timeout, causing stale-response loop
  https://github.com/openclaw/openclaw/issues/25272
  This is the closest if your Telegram group started doing typing/reactions, then no useful reply, especially around /compact / large context. It specifically mentions Telegram, compaction timing out, aborts during compaction, and loops.

• #44610, Session stops processing new messages after prompt-error: aborted
  https://github.com/openclaw/openclaw/issues/44610
  This is the closest if the group basically got stuck after an abort, where new messages stop getting processed correctly.

• #6301, sendPolicy deny rules bypassed by system notifications (abort, etc.)
  https://github.com/openclaw/openclaw/issues/6301
  Not the root cause, but very relevant to the “⚙️ Agent was aborted.” showing up in the group part.

• #64250, Context limit exceeded, I’ve reset our conversation to start fresh

https://github.com/openclaw/openclaw/issues/64250
Related if your setup got much worse after a recent update and now context handling feels broken.

My guess:
You might be hitting a combo of:

1. a Telegram compaction / timeout / abort bug, and
2. the abort notification getting surfaced back into the group.

So, not exactly one known issue, but very plausibly an already-known class of bug.

If you paste:

openclaw --version
node --version
openclaw logs --follow

and catch one failing group turn, I can tell you which of those issues it matches most closely.
If the logs show stuff like Unsubscribed during compaction, prompt-error: aborted, or repeated compaction retries, then it’s basically issue #25272 / #44610 territory.

[halcyon pilot]

Any issues or commits or PRs that have solved these problems?

[prisma birch]

I’m checking both the repo history and related PR titles, because some of these are fixed on main before the issue thread is updated.

I found at least a few actual fixes on main for the compaction side. I’m checking whether there’s a merged fix for the abort-spam / stuck-session side too.

Yeah, some of this has definitely been worked on.

Most relevant merged fixes I found

• PR #46889 / commit f77a684131
  https://github.com/openclaw/openclaw/pull/46889
  Makes compaction timeout configurable via agents.defaults.compaction.timeoutSeconds, and raises the default timeout. This is very relevant to the /compact still blows up / death spiral side.

• commit 6a458ef29e
  https://github.com/openclaw/openclaw/commit/6a458ef29e150b42e1f55ff8e01722728b11b218
  fix: harden compaction timeout follow-ups
  Also looks directly relevant to compaction timeout weirdness.

• commit cd4e2023ab
  https://github.com/openclaw/openclaw/commit/cd4e2023ab7279e2c1a618a94b963de0343cd7e1
  fix(agent): capture compaction retry AbortError for model fallback
  Relevant if aborts during compaction were bypassing recovery/fallback.

• PR #56343 / commit dd61171f5b
  https://github.com/openclaw/openclaw/pull/56343
  fix: prevent Telegram polling watchdog from dropping replies
  This one is a really good match for Telegram shows typing / activity, then no actual reply.

• PR #62600 / commit adb7b0d5d6
  https://github.com/openclaw/openclaw/pull/62600
  fix: compaction after tool use abortion cause agent infinite loop calls
  Relevant if abort + compaction pushed the session into bad looping behavior.

• commit 550deb8c18
  https://github.com/openclaw/openclaw/commit/550deb8c18909b00e28a7c12ac618da6a7869cf8
  fix: sweep stale chatRunState buffers for stuck runs
  Not Telegram-specific, but relevant to sessions getting wedged.

One important non-fix

• PR #25103 / commit 4b316c33db
  https://github.com/openclaw/openclaw/pull/25103
  This expanded / normalized abort trigger matching (stop, multilingual variants, etc.).
  So for the “⚙️ Agent was aborted.” spam, this is not a fix. If anything, it widened what counts as a stop trigger.

Also relevant

• commit 3a93a7bb1e
  https://github.com/openclaw/openclaw/commit/3a93a7bb1e47070a010cc190d005daa2ada8f0e5
  fix(security): enforce auth for abort triggers and models
  This helps make sure not just anyone can trigger abort shortcuts.

What I did not find quickly

• I didn’t find an obvious merged fix for the exact abort notification leaking into the group issue you asked about earlier (⚙️ Agent was aborted. showing up there).
• I also didn’t find a clean “fixes #44610” style commit for the exact prompt-error: aborted then session stays stuck case.

My read

If you’re on an older build, there’s a decent chance updating already helps, especially for:

• Telegram replies silently disappearing
• compaction timeout loops
• abort/compaction recovery

If you paste openclaw --version, I can tell you whether you’re before or after the likely fix range.

[halcyon pilot]

Already on 2026.4.12

[prisma birch]

Got it. If you’re already on 2026.4.12, then you’re past most of the fixes I found.

So my honest read is:

• this is probably not just “you’re on an old broken build”
• it looks more like a still-open edge case or a fresh regression
• the most suspicious buckets now are still:
  • Telegram group reply path gets wedged after abort/compaction
  • abort notification leaks back into the group
  • current context/compaction regression on 2026.4.x

So at this point I’d stop assuming “update fixes it” and move to logs + bug report.

What would make a strong issue

Include:

• OpenClaw: 2026.4.12

• Channel: Telegram group
• DMs still work
• group shows reaction / typing but no reply
• /think high hit Context limit exceeded
• /compact did not recover it
• after /stop or /new, group started sending ⚙️ Agent was aborted.

And attach logs from one bad turn:

openclaw logs --follow

Best signals to capture:

• Unsubscribed during compaction
• prompt-error: aborted