---
title: SKILL.md Validator — Free open-spec validator
description: SKILL.md is an open standard for portable, structured AI workflow specifications. The Cromus SKILL.md validator is free, deterministic, and runs in your browser — no signup, no API key, no LLM calls.
canonical: https://cromus.ai/validator/skill
source_html: https://cromus.ai/validator/skill
---

# SKILL.md Validator

> Free, deterministic validator for the SKILL.md open specification.

Paste or upload any SKILL.md file. Get a structured validation report against the open spec — including frontmatter, model governance, IO contract, fallback policy, and token estimates. No signup, no API key, no LLM calls.

---

## What SKILL.md is

SKILL.md is an open standard for portable, structured AI workflow specifications. A SKILL.md file captures everything a runtime needs to execute a capability safely:

- **Frontmatter** — name, version, model tier, ecosystem tier
- **Execution profile** — token estimates, complexity, latency budget
- **Model requirements** — minimum tier, preferred channels, fallbacks
- **Input/output contract** — typed schemas, required fields
- **Governance constraints** — references to ETHOS.md and MEMORY.md
- **Wrapper prompts** — per-platform adapters (OpenAI, Claude, Gemini)

The spec is public and lives at [github.com/cromus-ai/specs](https://github.com/cromus-ai/specs).

---

## What the validator checks

The Cromus validator runs ~40 deterministic checks per file across:

- **Schema conformance** — required frontmatter fields, valid enums, type correctness
- **Model governance** — tier declared, fallback chain present, cost mode set
- **IO contract** — input schema valid, output schema valid, no missing required fields
- **Token budget** — declared budget matches step estimates within tolerance
- **Portability** — wrapper prompts present for declared target platforms
- **Reference integrity** — referenced ETHOS.md and MEMORY.md fields exist

Every check is deterministic. The same file always produces the same report.

---

## Why a portable upstream validator matters

Most AI platforms ship their own opinionated spec (OpenAI Assistants config, Anthropic agent JSON, Google Interactions API, etc.). SKILL.md sits **upstream** of all of them — one portable source of truth that compiles down to any runtime config.

Validating against the open spec means:

- No vendor lock-in to a single runtime's schema
- Workflows are portable across LangGraph, CrewAI, AutoGen, n8n, Ollama, Claude Code, Claude Managed Agents, Google AI Studio, Replit AI, Open WebUI
- Teams catch structural errors before any runtime sees the workflow

---

## How to use

- **Web** — paste / upload at [cromus.ai/validator/skill](https://cromus.ai/validator/skill)
- **MCP** — call `validate_skill` on the Cromus MCP server at `https://mcp.cromus.ai/mcp`
- **CI** — same MCP tool, zero LLM calls, no-retention contract, safe for production CI

---

## Related

- [SKILL.md spec on GitHub](https://github.com/cromus-ai/specs)
- [ETHOS.md Validator](/validator/ethos.md)
- [MEMORY.md Validator](/validator/memory.md)
- [Pre-Execution Governance](/pre-execution-governance.md)