Skill Format (SKILL.md)

Quivly skills can be authored as code, in the Anthropic Agent Skills format, and imported from GitHub. A skill is a folder containing a SKILL.md file.

The file

---
name: renewal-risk-read
description: Assess renewal risk for a customer from health, usage, and conversations.
---

You are preparing a renewal risk assessment...

## What to produce
1. A risk rating with reasoning
2. Evidence from calls and tickets
3. Recommended next steps

How fields map into Quivly

SKILL.md	Quivly	Notes
`name`	Slash command + display name	Required. Lowercase-with-hyphens; falls back to the folder name. Max 64 chars
`description`	Description	Required. Max 1,024 chars
Markdown body	Instructions	Max 50,000 chars
`license`, `compatibility`, `metadata`	Preserved as source metadata	Optional spec fields
`allowed-tools`	Not imported	Anthropic tool names don’t map to Quivly’s catalog — the import flow suggests Quivly tools instead

Imported skills always land as drafts with no tools attached (or the tools you accepted from the AI suggestion step). Review, attach tools, and publish in Settings → AI → Skills.

What isn’t ingested

Only SKILL.md content is imported. Supporting folders from the spec (scripts/, references/, assets/) are not ingested; skills that have them are flagged so you know part of the skill lives outside Quivly. Files over 1 MB import with empty instructions.

FAQ

Can I use skills published for Claude?

Yes — any repo following the Agent Skills spec imports, including public community repos. You’ll attach Quivly data tools during import since Claude tool names don’t carry over.

Do I have to use GitHub to create skills?

No — most teams create skills in the app. The format matters when you want skills version-controlled in a repo or shared across tools.

MCP Tool Reference Import Skills from GitHub

​The file

​How fields map into Quivly

​What isn’t ingested

​FAQ

The file

How fields map into Quivly

What isn’t ingested

FAQ