data-chatbots

What problem does it solve?

This Skill eliminates the high risk of AI agents accidentally corrupting authoritative structured data (such as conference schedules, CRMs, or configuration stores) during automated edits, a problem that has caused costly production incidents including silent cancellation of confirmed talks, duplicate record creation from unique ID collisions, and unrecoverable data loss from unapproved overwrites.

Core Features & Use Cases

• Draft→apply workflow design: Enforces that AI only proposes changes for explicit human approval, never writes directly to canonical state, eliminating unvetted automated edits.
• Production-grade guardrails: Implements proven fixes for common failure modes including time-of-check-time-of-use (TOCTOU) overwrites, unique external ID deduplication, destructive in-place replacement prevention, and immutable audit logging for forensic recovery.
• Cross-surface parity: Includes patterns for web and Slack interfaces, FIFO client request queuing for long agent loops, LLM tracing with session grouping, and comprehensive test matrices for acceptance testing.
• Use case example: Use this Skill to build a conference scheduling copilot that drafts talk placements, speaker assignments, and cancellations for organizer approval, with safeguards to avoid accidentally overwriting confirmed bookings.

Quick Start

Use the data-chatbots skill to design a conference scheduling copilot that drafts talk placements and cancellations for human approval, with built-in guardrails to prevent accidental overwrites of confirmed bookings.