Skills Manifest — agentgrade

What is a Skills manifest?

A Skills manifest (/skills.json) lists the specific actions agents can take on your service. While OpenAPI describes your full API, skills.json highlights the key actions with their auth requirements — a curated menu for agents.

Example

Serve at /skills.json:

[
  {
    "name": "search",
    "description": "Search for items by keyword",
    "method": "GET",
    "path": "/api/search",
    "auth": "none"
  },
  {
    "name": "create",
    "description": "Create a new item",
    "method": "POST",
    "path": "/api/items",
    "auth": "x402"
  },
  {
    "name": "analyze",
    "description": "Run analysis on a dataset",
    "method": "POST",
    "path": "/api/analyze",
    "auth": "x402",
    "price": "$0.10"
  }
]

Key fields per skill

name: Short identifier for the action
description: What the skill does — agents read this to decide when to use it
method: HTTP method (GET, POST, etc.)
path: The endpoint URL
auth: Authentication required ("none", "x402", "bearer", etc.)

Why skills.json vs OpenAPI?

OpenAPI is comprehensive but verbose. Skills.json is a curated subset — the actions you *want* agents to take, with clear auth info. Think of it as the "quick start" vs the full API reference.

Verification

agentgrade checks that:

The file exists and contains valid JSON

Each skill has name and path fields

The declared paths actually respond (not 404)

Spec maturity

Informal convention. No formal standard exists. Multiple platforms define their own skill manifest formats (Alexa, Copilot, etc.). The /skills.json format used here is a simple, framework-agnostic convention.