ai-product-manager-master

What problem does it solve?

This Skill helps you turn an AI product idea into a production-ready plan by guiding you through scoping, choosing the right approach (workflow vs agent), designing evaluations, and running iteration loops based on failure modes—not vibes.

Core Features & Use Cases

• Agentic Protocol for product decisions: Diagnoses model capability shifts, eval/error-analysis maturity, workflow vs agent fit, token-cost constraints, team role boundaries, and reliability failure modes.
• Practical playbook rules: Enforces “error analysis first”, “binary pass/fail with calibration”, “pass^k for agent ship gates”, and “start cheapest then upgrade complexity”.
• Workflow walkthroughs (SOPs): Scoping → POC → build evals → eval-driven iteration → agent shipping → data flywheel.
• CLI execution layer: Bash scripts that operationalize the protocol and workflows into interactive checklists and decision reports.

Example use case: You want an LLM feature for RAG search quality improvements—this Skill helps you scope the real failure modes, set success criteria, design evals (golden set + judge calibration), and decide whether you need prompt/RAG changes or a deeper architectural shift.

Quick Start

Ask an AI assistant to “Act as ai-product-manager-master: run the Agentic Protocol for my AI PM problem, produce scoping guidance and an eval-driven iteration plan with failure modes and ship criteria.”