We Open-Sourced Our FinOps Brain. Here's Why.
Updated:
· 3 min read

We Open-Sourced Our FinOps Brain. Here's Why.

The FinOps Foundation framework transcribed for AI agents. One install command gives any LLM structured access to all capabilities, domains, personas, FOCUS spec, and waste sensors.

Table of contents

Open Table of contents

The Problem

LLMs know about FinOps the way they know about most professional frameworks — broadly, generically, and without the structured detail that makes advice grounded.

Ask an LLM to “optimize my AWS costs” and you’ll get reasonable-sounding advice: right-size instances, use Savings Plans, implement tagging. All correct. All vague.

The Foundation publishes a framework with specific capabilities, maturity criteria, functional activities by role, and measures of success. That structure exists in public GitHub repos. But LLMs don’t have it loaded as working context — they hallucinate a plausible-sounding version instead.

What’s in the Skill

What’s transcribed (from Foundation repos)

Every reference file traces to a specific upstream source. No custom content in the reference directory.

From finopsfoundation/framework:

  • All capabilities — each with definition, Crawl/Walk/Run maturity criteria, functional activities by persona, measures of success, and inputs
  • All domains — with descriptions and capability mappings
  • All personas — with primary goals, objectives, frustrations, key metrics, and FinOps benefits

From FOCUS_Spec:

  • Specification overview, design principles, conformance rules
  • Cost and Usage dataset with all column definitions
  • Glossary of terms

From finopsfoundation/kpis:

  • KPI definitions from the community
  • Waste sensor definitions — standardized IDs, descriptions, and methodology for calculating savings opportunities and waste percentages

What’s custom (just the routing)

SKILL.md is the only custom file. It contains:

  • Expert advisor prompt — persona adaptation, quality standards, engagement modes
  • Business problem routing table — maps user questions to relevant Foundation capabilities
  • Condition-keyed loading — detects what the user is asking about and loads only the relevant reference files

This means the skill loads a couple of files for a targeted question, not the whole knowledge base. “How should I structure Savings Plans?” loads the commitment discounts capability. “Full maturity assessment” loads all capabilities systematically by domain.

Try It

Install in any project:

npx skills add suan-digital/cloud-finops

Then try:

  • “Assess our FinOps maturity” — walks through capabilities using Crawl/Walk/Run criteria
  • “How should I structure Savings Plans?” — routes to commitment discounts capability
  • “We can’t attribute costs to teams” — routes to cost allocation + shared costs
  • “Compare costs across our AWS and GCP accounts” — routes to FOCUS spec + data normalization

GitHub: suan-digital/cloud-finops


For a tailored FinOps assessment or advisory engagement: suan.digital/contact

AI-assisted drafting, human-reviewed and edited.