hatch3r-ai-feature

What problem does it solve?

Prevents unreliable LLM feature launches by forcing teams to define measurable success, run evals, and verify production quality with telemetry and fallback behavior.

Core Features & Use Cases

• Eval-driven workflow for AI features: turns “good enough” prompts into repeatable, thresholded experiments using a golden dataset and iterated prompt versions.
• Production readiness gates: wires cost telemetry, cache control, hallucination/refusal SLI checks, and CI blocking when metrics regress.
• Resilience via fallback chain: ensures graceful degradation through a primary-to-secondary-to-static fallback strategy with circuit-breaker and retry primitives.

Quick Start

Run this skill before deploying a new LLM-powered feature by first asking it to identify ambiguity in your inputs and then proceed through writing golden evals, prompts, thresholds, eval runs, and production verification.