Skip to content
SkillVault logo SkillVault

Skill import guide

Every SkillVault skill downloads as a standard .aiskill file (JSON or YAML). This guide shows you exactly where each part goes on your platform of choice.

Anatomy of an .aiskill file

  • prompts.system — the system prompt that defines your agent's role and rules.
  • prompts.user_template — your recurring message, with {{variables}} to fill in.
  • variables — each placeholder documented with description, required flag and example.
  • install / usage — platform install steps and usage flow.

Step 1 — Download and open the file

Download the .aiskill file from any skill page (sign in required). It is plain JSON — open it in any editor, or import it into the online editor for tweaking. YAML exports work the same way.

Step 2 — Install on your platform

ChatGPT GPTs

  1. Open ChatGPT → Explore GPTs → Create.
  2. Paste prompts.system into the Instructions field.
  3. Name the GPT after the skill and save (Plus/Team plans required for publishing).
  4. Start new chats with the filled user_template.

Claude Code

  1. Save the file in your repo, e.g. .claude/skills/my-skill.json.
  2. Or create a slash command at .claude/commands/my-skill.md containing the system prompt.
  3. Reference the user template in your task message.
  4. Run claude and invoke the command with /my-skill.

Cursor

  1. Create .cursor/rules/my-skill.mdc in your project.
  2. Add frontmatter with description and alwaysApply: false.
  3. Paste the system prompt as the rule body.
  4. Trigger it in chat or agent mode by asking for the skill; keep the user template handy.

Dify

  1. Create a new Agent (or Chatflow) app.
  2. Paste prompts.system into the orchestrator's prompt panel.
  3. For each variable in variables, add an input field with the same name.
  4. Publish and test with a filled user template.

LangGraph

  1. Load the JSON at graph-build time.
  2. Use prompts.system as your system message.
  3. Render prompts.user_template with a simple str.format-style substitution of the variables.
  4. Optionally branch on the skill's category/tags in your graph state.

Step 3 — Fill the template correctly

Replace every {{variable}} with real content. A filled example:

example-user-message.txt 
Review the following {{language}} code.

Focus areas: {{focus_areas}}

```{{language}}
{{code}}
```

Check the skill page's Variables table for which fields are required and what good inputs look like.

Troubleshooting

  • The agent ignores the skill — the system prompt is probably truncated; platforms cap instruction length, so prefer focused skills.
  • Literal {{variable}} in responses — a placeholder wasn't filled before sending.
  • JSON parse errors — re-download the file or open it in the editor and hit Format.