← Prompt library
Agent ops & evalsFree→ Claude / ChatGPT

Prompt/Model Change Release Plan

Plan the eval gate, canary and rollback for a prompt or model change.

What you get: A staged release plan with success metrics, thresholds and rollback triggers.

input
output — Claude / ChatGPT
I'm changing {{what_is_changing}} (prompt / model / tool schema / retrieval) for an agent that {{agent_description}}. Traffic: {{traffic}}. Risk if it regresses: {{risk}}.

Create the release plan: the exact eval suite to run and the pass thresholds for quality, safety and cost; a shadow-mode phase (what to compare and for how long); canary percentages and hold times; the metrics and numeric rollback triggers (zero-tolerance metrics first); how rollback works and how long it takes; who is on watch and for how long after full rollout; the changelog entry; and the documentation that must be updated. Call out hidden coupling (other prompts, tools or indexes affected by this change).