Skip to main content
Skills are reusable prompt modules that extend Octo’s capabilities. Each skill is a directory containing a SKILL.md file with YAML frontmatter.

SKILL.md Format

---
name: summarize-pdf
description: Summarize PDF documents into concise bullet points
dependencies:
  python:
    - pymupdf
  system:
    - poppler
---

When asked to summarize a PDF:

1. Use the `Read` tool to read the PDF file
2. Extract the key sections and main arguments
3. Produce a bullet-point summary with:
   - Main thesis / purpose
   - Key findings (3-5 points)
   - Conclusions and recommendations
4. Keep the summary under 500 words unless asked for more detail

Frontmatter Fields

FieldRequiredDescription
nameYesSkill identifier (lowercase, hyphens ok)
descriptionYesOne-line description shown in /skills list
dependenciesNoDependencies to auto-install

Dependency Types

dependencies:
  python:
    - pymupdf>=1.23.0
    - pandas
  npm:
    - "@playwright/mcp"
  mcp:
    - name: playwright
      type: stdio
      command: npx
      args: ["@playwright/mcp@latest"]
  system:
    - poppler       # shown for manual install
    - ffmpeg
TypeAuto-installed?How
pythonYespip install into active venv
npmYesnpm install --prefix .octo/
mcpYesAdded to .mcp.json
systemNoDisplayed for manual installation

Installation Location

Skills are installed to .octo/skills/<name>/SKILL.md.

Invoking Skills

Once installed, invoke a skill as a slash command:
/<skill-name>
The skill’s prompt is injected into the conversation, guiding the agent’s behavior for that interaction.

Publishing Skills

Skills can be shared via Git repositories. The marketplace commands (/skills search, octo skills search) discover published skills.
Keep skills focused on a single task. A skill that does one thing well is more reusable than one that tries to do everything.