Bug: memory tool response leaks into Telegram output
Description
When the memory tool is called (add/replace/remove), its internal response — including the old_text search string and operation indicators (+/-/~ prefixes) — is being rendered directly into the public Telegram message output.
Example of what appears in Telegram (user-visible):
🧠 memory: "+memory: "BUG: memory tool response...""
🧠 memory: "~memory: "Matrix homeserver UR""
🧠 memory: "-memory: "OpenClaw models via""
These are internal tool response artifacts that should never reach the user.
Frequency
Seen multiple times by the same user.
Environment
- Platform: Telegram
- Hermes gateway with user-mode systemd service
Expected behavior
The memory tool should operate silently. Only the final confirmation (if any) should reach the user. The old_text string, operation markers (+/-/~), and internal state should never appear in the chat.
Possible cause
The memory tool's response formatter may be passing raw operation metadata into the Telegram rendering layer instead of suppressing it for messaging platforms.
Bug: memory tool response leaks into Telegram output
Description
When the
memorytool is called (add/replace/remove), its internal response — including the old_text search string and operation indicators (+/-/~ prefixes) — is being rendered directly into the public Telegram message output.Example of what appears in Telegram (user-visible):
These are internal tool response artifacts that should never reach the user.
Frequency
Seen multiple times by the same user.
Environment
Expected behavior
The memory tool should operate silently. Only the final confirmation (if any) should reach the user. The old_text string, operation markers (+/-/~), and internal state should never appear in the chat.
Possible cause
The memory tool's response formatter may be passing raw operation metadata into the Telegram rendering layer instead of suppressing it for messaging platforms.