What problem does it solve?
When users are frustrated with an agent's behavior or encounter problems with Letta Code, there is no structured way to capture that context and deliver it to the Letta team. This Skill turns conversational frustration into a factual, evidence-backed feedback report.
Core Features & Use Cases
- Consent-gated submission: Asks the user before submitting and never sends feedback merely because frustration was expressed.
- Structured report writing: Guides the agent to write a factual report including agent name, user context, the task underway, expected versus actual behavior, and concrete evidence such as error text and reproduction steps.
- CLI-based delivery: Submits the report with the
letta feedback --message command, which automatically attaches agent and conversation identifiers.
- Use Case: A user reports that the agent repeatedly failed a refactoring task. The agent asks for consent, compiles the error output and failed commands into a report, and submits it to the Letta team.
Quick Start
Ask the agent to submit feedback about the problem you just experienced to the Letta team.