Skip to main content
A skill is a markdown playbook the Surfer MCP server ships. Your assistant reads it and executes it with the server’s tools, pausing where your input matters. You ask for the outcome and the playbook carries the steps you would otherwise prompt one by one. Over MCP, playbooks update when the server updates, with no separate installation. You can also browse the source or install a local copy from the skills repository.

How to use a skill

You need no special syntax. A skill triggers from a natural request: The assistant matches the request to the surfer-write-article skill and runs the playbook. You can also name the skill: The assistant lists the available skills, each with its name and a description of when it applies, then fetches the playbook it picked.
Skills served over MCP are not slash commands. The Surfer MCP server registers tools only, and no MCP prompts. Your client does not show them in its slash command list or prompts menu. Ask in plain language or name the skill in your message.Locally installed skills can also appear as slash commands, depending on your client. For example, in Claude Code, run /surfer-write-article after installing it from GitHub.
If your assistant cannot find any skills, its cached tool list is stale; Troubleshooting has the refresh step for each client.

Available skills

Which skill do I need?

  • Nothing written yet and you want a finished draft: .
  • Nothing written yet and you want a plan for a human writer: (adds SEO Guidelines and AI Search facts) or (headings only).
  • Content already exists: .
  • Setting standards for future content: .
  • You do not know what to work on: . It lists the workspace’s recommendations, lets you pick one, and hands the work to surfer-optimize-content or surfer-write-article.
  • Setting a writing style or tone profile: no skill. That is a custom voice, not a template, so manage voices with the voice tools.
  • One-off direction for a single article: no skill either. It goes in the custom instructions that surfer-write-article collects.
Use a skill by default for the six workflows above. Drop to free-form prompts for what the playbooks do not cover: AI Tracker reads, reporting, workspace administration, or a single tool call. Use cases has example prompts for those.

What happens when you run one

This is what the surfer-write-article playbook does, stage by stage. The playbook’s own numbering groups the work slightly differently; ask your assistant to show you the playbook for the exact wording.
1

Confirms the setup before spending a credit

The assistant verifies the workspace you supplied or asks which one to use when several are active. For a new editor, it collects:
  • The main keyword, plus up to 19 secondary keywords
  • The location (default United States) and device (default mobile)
  • Whether to apply your brand knowledge
  • One template, a Surfer preset or one of your workspace’s own
  • A voice
  • Custom instructions
  • Any word count or score targets
If you are continuing from a recommendation, outline, or brief, the assistant verifies and reuses that Content Editor. It keeps the existing settings unless you request a change. A request for a separate article creates a new editor.Creating a Content Editor uses one Content Editor credit, and the playbook makes the create request safe to retry (Credits and limits).
2

Waits for SERP analysis

Creating the Content Editor takes time; the assistant waits for it or checks its progress, whichever your client supports. It reports a failure or a bounded timeout together with the editor id.
3

Reviews the content plan

It reads three things:
  • The SEO Guidelines: terms, structure targets, topics, questions, competitors
  • The AI Search facts, when AI Search is a goal
  • The editor itself, to confirm the brand toggle, template or voice, and instructions took effect
If you ask for different competitors, it updates the competitor list and re-reads the guidelines before writing.
4

Pauses for your outline approval

When you ask for a reviewable outline, the assistant starts the article in manual outline mode. Generation pauses once the outline is drafted, and the assistant fetches it and shows it to you. Enterprise and some older plans reserve an AI Article credit when it starts; other plans include AI Articles in the Content Editor. See Credits and limits. Only the version you approve goes back, and nothing is written until you sign off.
5

Generates or continues the article

It starts a new draft or resumes the article already in progress. If a response is lost, it checks for the article before taking further action. It reports an unclear outcome and stops.After generation completes, it reads the stored draft and its SEO, AI Search, and total scores. If the AI Search score shows error or unavailable, see Troubleshooting.
6

Iterates toward your targets

If you set an SEO or AI Search score target and it is unmet, the assistant revises in rounds. Each round:
  1. Revises the draft against the guidelines and facts.
  2. Replaces the editor’s content with the revision.
  3. Re-reads the version Surfer stored (see Content Editors).
  4. Waits for the scores to recalculate.
It stops after 3 to 5 rounds or when the score plateaus, and it does not chase the unified total.
7

Hands off

You get the content, the SEO, AI Search, and total scores separately, the Content Editor id, and an edit or share link to the editor.
Every skill follows the same pattern: confirm inputs, execute, pause where your judgment is needed, deliver with links back to Surfer. Which of the steps consume credits is on Credits and limits. Tools: workspaces, Content Editors, guidelines, AI Articles, templates and voices.

Inspect a playbook

Ask your assistant to list the available skills or show you a playbook before running it. Reading a playbook spends no credits. You can also read each playbook’s SKILL.md in the skills repository.

Get the skills from GitHub

surferseo/skills is the source repository for the playbooks bundled with Surfer MCP. The skills are MIT licensed, so you can inspect and adapt them. Install a local copy if your client does not load skills from the server or you want to control when your playbooks update. Local skills still need a connected Surfer MCP server. The repository also includes surfer-connect for setup and connection problems. It is available through local installation; the MCP server supplies the six workflow skills listed above.
1

Install the skills

With Node.js 22.20 or newer, run the skills CLI from your project directory to install every Surfer skill, including connection setup:
To install one workflow, replace '*' with its name, such as surfer-create-outline. Add -a codex or -a claude-code to target one assistant.
2

Connect Surfer

Ask your assistant to use surfer-connect, or follow the setup guide for your client. If Surfer is already connected, you can use the workflow skills immediately.
3

Update your local playbooks

When you want the latest versions, run this command from your project directory to update installed skills:
Local copies update through the CLI. Skills served over MCP update with the server deployment.