What problem does it solve?
When IMF Research Assistant (RA) Skills fail or produce unsatisfactory results, users have no standardized, privacy-safe way to share feedback with the development team. This Skill eliminates ad-hoc error reporting by providing a structured, consent-first workflow to capture relevant context for debugging, without unauthorized data collection or remote uploads.
Core Features & Use Cases
- Dual Report Scenarios: Handles both system/execution errors (crashes, timeouts, missing output files, SDK failures) and unsatisfactory answers after repeated workflow attempts.
- Strict Consent Guardrails: Never creates reports silently; only generates files after explicit user approval, with a maximum of 5 reports per conversation session to avoid spam.
- Structured Debug Context: Captures relevant details including failed commands, error output, user request context, and environment information in a standardized local JSON format for the development team.
- Use Case Example: If a user tries to fetch WEO data for 10 countries and gets a timeout after 3 retries, they can consent to a report that includes the failed API calls, error messages, and their original request to help the team fix the underlying issue.
Quick Start
Ask the AI to prepare a local error report for a recent visible RA-Skills failure, then provide optional context about what went wrong when prompted.