openclaw - 💡(How to fix) Fix suppressToolErrors does not work for mutating tools (message/exec) [1 participants]

Every failed message tool call pushes a ⚠️ warning to the user via the channel, even if the agent handles the error gracefully. When retry logic causes 3+ failures, the user sees 3+ warnings. Skill-level error recovery (feishu-message-error-recovery) that switches methods after 1 failure instead of retrying.

TL;DR

Implementing a flag like agents.<id>.directives.suppressMutatingWarnings could help suppress warnings for mutating tools.

Guidance

• Review the code at pi-embedded-runner-BFOhIiuh.js:980 to understand why suppressToolErrors is not working for mutating tools like message.
• Consider adding a per-agent flag to control warning behavior for mutating tools, as suggested in the issue.
• Evaluate the current workaround (feishu-message-error-recovery) and assess whether it can be improved or replaced with a more robust solution.
• Investigate how suppressToolErrors is currently implemented and why it does not apply to mutating tools, to inform the design of a new flag or fix.

Example

No code example is provided due to lack of specific implementation details in the issue.

Notes

The suggested fix implies a need for a more fine-grained control over warning behavior, particularly for mutating tools. The current workaround may not be ideal, as it involves switching methods after a single failure.

Recommendation

Apply workaround: Implement a per-agent flag like agents.<id>.directives.suppressMutatingWarnings to control warning behavior for mutating tools, as this seems to be a targeted solution to the problem described.