Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

After upgrading to 2026.5.20, Telegram shows user-visible technical banners like "⚠️ 🛠️ ... failed" for shell commands where grep exits with code 1 due to no matches (normal grep behavior).

Steps to reproduce

1. Run OpenClaw 2026.5.20 in Telegram group/topic.
2. Let an agent execute a shell pipeline that can return grep exit 1 on no match, for example:
   ps -eo pid,ppid,stat,cmd | tail -n +1 | grep -E 'find|grep|xargs|bash -lc' | grep -v 'ps -eo' | grep -v grep
3. Observe Telegram output after tool completion.
4. Command returns exitCode: 1 (no matches), but chat shows "⚠️ 🛠️ ... failed".

Expected behavior

Normal non-critical command outcomes (such as grep no-match exit 1) should not be surfaced as alarming user-facing failure banners in Telegram. Internal tool-status noise should be suppressed unless explicitly in full debug/verbose mode.

Actual behavior

Telegram displays a user-visible technical failure banner:
"⚠️ 🛠️ failed"
for a command that ended with exitCode: 1 because grep found no matches. This appears as a runtime/tool failure to end users even when it is a normal grep outcome.

OpenClaw version

2026.5.20

Operating system

Ubuntu 24.04 on VPS (Hyperforce), client macOS 15.6

Install method

npm global + OpenClaw gateway service

Model

gpt-5.5 (openai-codex)

Provider / routing chain

openclaw -> @openclaw/codex -> telegram channel dispatch

Additional provider/model setup details

Core pinned to 2026.5.20. Plugins pinned to compatible versions (@openclaw/[email protected], @openclaw/[email protected]). Reproduced in Telegram topic sessions. No API keys included.

Logs, screenshots, and evidence

Observed banner in Telegram: ⚠️ 🛠️ ps -eo pid,ppid,stat,cmd | tail -n +1 | grep -E 'find|grep|xargs|bash -lc' | grep -v 'ps -eo' | grep -v grep ... failed
Trajectory/session evidence shows command exitCode: 1 (grep no match). Source review between v2026.4.23 and v2026.5.20 indicates expanded visible tool-progress pipeline and later partial suppression fixes.

Impact and severity

Affected: Telegram users in group/topic chats.
Severity: Medium (high UX noise; appears as real failure to users).
Frequency: Intermittent but reproducible for grep no-match paths.
Consequence: Confusion, reduced trust in agent replies, unnecessary incident triage.

Additional information

Last known good behavior baseline for this UX: pre-5.x / older pipeline (no visible failed-tool banners for this pattern).
First observed bad behavior: after upgrade to 2026.5.20.
This report requests upstream fix; no local workaround requested.