Skip to content

SKILL.md

Latest commit

 

History

History
119 lines (101 loc) · 5.52 KB

SKILL.md

File metadata and controls

119 lines (101 loc) · 5.52 KB
Raw
name mcp-foundation
description Build the agent-native foundation layer — protocol files, API endpoints, agent profiling, referral beacons, feedback loop — for any web service that needs to be natively consumable by AI agents. Each feature in this skill is a portable, config-driven generator that ships a single check against the Cloudflare Agent Readiness scanner. Install, point at a config, get a 100% score. Part of Mahoney Context Protocol.

mcp-foundation

Part of Mahoney Context Protocol.

Build agent-ready protocol layers for any site, scoring 100% on Cloudflare's Agent Readiness scanner.

Each feature is a portable generator: takes a config, emits a protocol file. Compose them all to ship the full machine layer.

Features (incremental — added one at a time)

  • Feature 01: robots.txt generator
  • Feature 02: sitemap.xml generator
  • Feature 03: Link headers (RFC 8288)
  • Feature 04: Markdown content negotiation
  • Feature 05: Content Signals validator + ai-preferences manifest
  • Feature 06: API Catalog (RFC 9727 + RFC 9264)
  • Feature 07: OAuth/OIDC discovery (RFC 8414 + OIDC Discovery 1.0)
  • Feature 08: OAuth Protected Resource (RFC 9728)
  • Feature 09: MCP Server Card (SEP-1649 + SEP-1960 + SEP-2127)
  • Feature 10: A2A Agent Card (canonical + legacy paths)
  • Feature 11: Agent Skills index (v0.2.0 + v0.1.0 + SKILL.md files)
  • Feature 12: WebMCP runtime (navigator.modelContext registration)
  • Feature 13: llms.txt site summary (https://llmstxt.org)
  • Feature 14: operate.txt (Mahoney Context Protocol operate-txt v1.0)
  • Feature 15: security.txt (RFC 9116 + MCP Agent-Disclosure extension)
  • Feature 16: JSON-LD structured data (Schema.org)
  • Feature 17: ai.txt (Spawning + AI Visibility v1.1.1) + humans.txt
  • Feature 18: OpenAPI discovery (index + canonical paths + Worker)
  • Feature 19: did:web verifiable identity (W3C DID Core v1.1)

Foundation status: 19/19 features complete. Coverage spans RFC 9116 / 9309 / 9727 / 9728, OpenID Connect Discovery 1.0, OAuth 2.0 Authorization Server Metadata, OpenAPI 3.x, W3C DID Core v1.1, Schema.org JSON-LD, and the Mahoney Context Protocol operate-txt v1.0 spec.

Feature 19 is the closer. Once an agent resolves did:web:<domain> and verifies the keys, every other foundation file becomes trust-rooted via the MahoneyContextProtocolFoundation service anchor. Custom service type is the third namespace marker (after operate.txt and the security.txt Agent-Disclosure: extension). Future move: submit to W3C DID Property Extensions registry.

Roadmap

  • Feature 01-19 shipped
  • Phase 2: build the it-IS-agent-ready scanner that consumes mcp-foundation outputs
  • mcp-runtime: WebMCP framework as composable modules
  • DID Property Extensions registry submission for MahoneyContextProtocolFoundation service type
  • AI training opt-out namespace consolidation spec (EU AI Act Article 53 framing)
  • Agent Security Disclosure spec (companion to security.txt Agent-Disclosure field)

Feature 18 ships OpenAPI discovery scaffolding — .well-known/openapi-index.json catalogs every spec the site serves, primary spec at /openapi.{json|yaml} canonical path, redirect from .well-known/openapi.{json|yaml} for the emerging convention, Link: describedby header advertising. mcp-foundation does not generate OpenAPI documents — it provides the discovery layer for ones the user already has.

Feature 17 ships clean reference implementations of two existing ai.txt formats (Spawning + AI Visibility Directory v1.1.1). Namespace fragmentation in AI training opt-out is a future consolidation opportunity (EU AI Act Article 53, Hamburg ruling Dec 2025) — not claimed here. humans.txt per humanstxt.org spec, with optional <link rel="author"> snippet for the HTML head.

Feature 16 is the first feature emitting per-page artifacts in addition to site-level files. Schema.org JSON-LD with an agent-readable SoftwareApplication block declaring the site as a machine-callable service.

Note: Features 01–12 cover Cloudflare's current Agent Readiness scanner rubric. Feature 13 is the first feature beyond the upstream scanner — the foundation now covers the rubric AND extends past it.

Feature 14 implements the Mahoney Context Protocol operate-txt v1.0 spec. Every emitted file cites the spec URL — claiming the namespace by shipping the reference implementation.

Feature 15 ships RFC 9116 with an MCP custom extension — Agent-Disclosure: — a dedicated channel for AI agents to report security issues they discover at runtime. The custom field is the seed for a future Mahoney Context Protocol companion spec on agent security disclosure.

Quick start

npm install
cp config.example.yml config.yml   # edit hostname + bot policy + urls
npm run generate -- config.yml     # writes ./output/{robots.txt,sitemap.xml}
npm test                           # unit tests
npm run test:verify -- https://your-site.example.com   # live scan (robots)
bash test/verify/sitemap-xml.sh https://your-site.example.com  # live scan (sitemap)

Drop output/robots.txt at the root of your site. Re-run the verify script against the deployed URL to confirm a passing score on the robots.txt portion of Cloudflare's Agent Readiness scanner.

Repository

https://github.com/MahoneyContextProtocol/mcp-foundation