caio-review

Interrogates AI plans on evals, risk classification, cost economics, and hiring decisions.

25.3k|3.6k|Updated Oct 19, 2025
One-click install
npx skills add https://github.com/alirezarezvani/claude-skills --skill caio-review
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: caio-review
Source: https://github.com/alirezarezvani/claude-skills/tree/main/c-level-agents/skills/caio-review
Command: npx skills add https://github.com/alirezarezvani/claude-skills --skill caio-review

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Shipping AI features without eval criteria, regulatory classification, or cost projections leads to failed launches, compliance exposure, and runaway vendor spend. This Skill applies a Chief AI Officer's six forcing questions to any AI plan before commitment.

Core Features & Use Cases

  • Eval Discipline Check: Blocks any AI feature that lacks a defined eval set, error-rate SLO, and fallback behavior.
  • Regulatory & Model Decisions: Classifies use cases under the EU AI Act and recommends API vs fine-tune vs build using companion calculator scripts.
  • Cost & Org Planning: Projects 12-month cost trajectories and identifies the correct next AI hire (AI engineer vs ML engineer vs research scientist).
  • Use Case: Before signing a multi-year LLM API contract, run the review to get a SHIP/SHARPEN/BLOCK verdict with TCO comparison and risk tier.

Quick Start

Ask the AI to run a CAIO review on your AI feature plan, covering eval criteria, EU AI Act risk tier, model build-vs-buy choice, and cost projection.

Frequently Asked Questions about caio-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I evaluate an AI feature before shipping it?

Define an eval set of 50-100 representative inputs with expected outputs or a grading rubric, set a quantified error-rate SLO, and specify a fallback such as human review. This Skill walks through all six CAIO questions and outputs a SHIP, SHARPEN, or BLOCK verdict.

Should I use an API, fine-tune, or build my own model?

Most B2B SaaS use cases fit a hosted API; fine-tuning suits domain-specific behavior with labeled data, an ML team, and high volume; building from scratch applies to under 1% of cases. The companion model_buildvsbuy_calculator.py script computes the economic breakeven for your workload.

How do I classify an AI use case under the EU AI Act?

Run the ai_risk_classifier.py script with a use case description to get a PROHIBITED, HIGH, LIMITED, or MINIMAL tier. HIGH tier triggers conformity assessment and EU database registration, while LIMITED requires transparency disclosures.

When does self-hosting an LLM become cheaper than an API?

Breakeven typically occurs at 1-10 billion tokens per month for 70B-class models, but self-hosting adds hidden costs for ops, monitoring, failover, and model updates. The ai_cost_economics.py script projects the 12-month trajectory for your specific workload.

What are the limitations of this AI plan review?

The review is a structured interrogation framework, not an automated audit; it depends on accurate inputs about your workload, data, and team. Regulatory outputs are guidance and do not replace legal counsel for EU AI Act conformity decisions.