openclaw 2026.6.5-beta.2: QQBot Thinking Scaffolding Stripped Before Delivery

OpenClaw’s latest beta release, 2026.6.5-beta.2, addresses a notable content leakage issue in the QQBot channel. Model reasoning and <thinking> scaffolding is now stripped before native delivery, preventing raw internal model outputs from appearing in channel replies.

The release also improves how MCP tool results are coerced — resource links, resources, audio, malformed images, and future non-text/image blocks are now handled more gracefully rather than causing delivery errors.

These changes follow the stable 2026.6.1 release (June 3), which focused on recovery from interrupted tool calls, stale session bindings, compaction handoffs, and media delivery retries across Telegram, WhatsApp, iMessage, Slack, Discord, and Microsoft Teams.

Users on the beta channel can update with openclaw update --channel dev.