Docfiy hosts a skill.md file at the root of your project that describes what AI agents can do with your product.
The skill.md specification is a structured, machine-readable format that makes capabilities, required inputs, and constraints for products explicit so that agents can use them more reliably.
Docfiy automatically generates a skill.md file for your project by analyzing your documentation with an agentic loop. This file stays up to date as you make updates to your documentation and requires no maintenance. You can optionally add a custom skill.md file to the root of your project that overrides the automatically generated one.
skill.md file can take up to 24 hours.View your skill.md by appending /skill.md to your documentation site's URL. Docfiy only generates skill.md files for documentation sites that are public.
Open the skill.md for this site.
Use skill.md files with agents
If you use a reverse proxy, configure it to forward /skill.md, /.well-known/skills/*, and /.well-known/agent-skills/* paths to your Docfiy subdomain.
When users connect to your MCP server, their agents can discover and use your skill.md files as MCP resources without installing the skills separately.
To add your skills to an agent's context manually, use the skills CLI.
This adds your product's capabilities to the agent's context so it can take actions on behalf of users.
skill.md structure
Docfiy generates a skill.md file following the agentskills.io specification. The generated file includes:
- Metadata: Project name, description, and version.
- Capabilities: What agents can accomplish with your product.
- Skills: Specific actions organized by category.
- Workflows: Step-by-step procedures for common tasks.
- Integration: Supported tools and services.
- Context: Background on your product's architecture.
Custom skill files
Add custom skill files to override the automatically generated skill.md. Docfiy supports hosting a single skill file and a directory for multiple skills. If you delete all custom skill files, Docfiy generates a new skill.md file.
Single skill file
Add a skill.md file to the root of your project to override the automatically generated file.
Multiple skill files
Add multiple skill files to the .docfiy/skills/ directory in your project. Each skill must be in its own subdirectory with a SKILL.md file:
When you have multiple skills, the /skill.md endpoint redirects to the /.well-known/skills/index.json discovery endpoint, which lists all available skills. The discovery endpoints make each skill individually accessible.
You can use both approaches together with a root skill.md file and .docfiy/skills/ directory. The discovery index includes all skills.
Use symlinks to avoid duplication
If your skill files live elsewhere in your repository (for example, in a plugins/ or skills/ directory), you can symlink .docfiy/skills to that location instead of duplicating files:
Docfiy resolves symlinks during deployment, so the skill files are discovered and served as if they were directly inside .docfiy/skills/. This works with both directory symlinks and individual skill symlinks.
Frontmatter fields
Custom skill.md files must start with YAML frontmatter.
| Field | Type | Description |
|---|---|---|
name | string | The name of your skill. |
description | string | A brief description of what your skill does. |
license | string | The license for your skill (for example, MIT or Apache-2.0). |
compatibility | string | Requirements or compatibility notes (for example, runtime dependencies). |
metadata | object | Additional metadata as string key-value pairs (for example, author or version). |
allowed-tools | string | Space-delimited list of pre-approved tools the skill may use (experimental). |
Skills discovery endpoints
Docfiy hosts skills directories at /.well-known/skills/ and /.well-known/agent-skills/ that agents can use to discover and fetch your skill files programmatically.
Agent-skills discovery (recommended)
The /.well-known/agent-skills/ endpoint follows the agent-skills 0.2.0 discovery spec and includes content integrity verification.
GET /.well-known/agent-skills/index.json returns a JSON manifest listing all available skills:
| Field | Description |
|---|---|
$schema | Schema URL for the agent-skills 0.2.0 discovery spec. |
name | A URL-safe slug derived from the name in your skill.md frontmatter. |
type | The skill format. Always skill-md. |
description | A brief description from your skill.md frontmatter, truncated to 1024 characters. |
url | The path to fetch the full skill file. |
digest | A sha256 hash of the skill file content for integrity verification. |
GET /.well-known/agent-skills/{name}/SKILL.md returns the skill.md file for a specific skill identified by its slugified name from the index.
Skills index
The /.well-known/skills/ endpoint is the original discovery format.
GET /.well-known/skills/index.json returns a JSON manifest listing all available skills:
The name field is a URL-safe slug derived from the name in your skill.md frontmatter.
Individual skill files
GET /.well-known/skills/{name}/skill.md returns the skill.md file for a specific skill identified by its slugified name from the index.
Agent card
Docfiy hosts an Agent-to-Agent (A2A) agent card at /.well-known/agent-card.json. The agent card is a standardized JSON document that helps A2A-compatible agents discover your documentation site and available skills in a single request.
GET /.well-known/agent-card.json returns a JSON document following the A2A agent card 0.3 schema. Each entry in the skills array corresponds to a skill from your skills discovery endpoints.
A2A-compatible agents fetch /.well-known/agent-card.json to discover your site by name and description, follow documentationUrl to retrieve human-readable content, and iterate over skills to fetch each skill.md file. The card also advertises a supportedInterfaces array so agents can negotiate transport before making a request.
| Field | Description |
|---|---|
protocolVersion | A2A protocol version. Always 0.3. |
preferredTransport | Default transport for clients that do not negotiate. Always HTTP+JSON. |
supportedInterfaces | Array of { url, protocolBinding, protocolVersion } entries describing how agents can reach your site. |
provider | { url, organization } identifying the documentation site. organization is the site title. |
defaultInputModes | Media types the agent accepts as input. Always ["text/plain"]. |
defaultOutputModes | Media types the agent produces. Always ["text/plain"]. |
capabilities | Feature flags. Currently { streaming: false, pushNotifications: false }. |
skills | The skills exposed at /.well-known/agent-skills/. |
URLs in the card (url, documentationUrl, provider.url, and each skill URL) use your configured custom domain, so the published card always advertises the canonical domain instead of the *.docfiy.com subdomain.
If you use a reverse proxy, configure it to forward /.well-known/agent-card.json to your Docfiy subdomain.
The agent card complements MCP by providing a lightweight discovery layer that does not require establishing a session.