Skip to main content

Prerequisites

  • The fastskill binary on your PATH (see Installation: Homebrew on macOS/Linux, Scoop on Windows, or release archives)
  • OPENAI_API_KEY when you use embedding search or reindex

Five-minute setup (CLI)

1

Install the CLI

Follow Installation, then confirm:
2

Create a manifest

Create skill-project.toml at your project root:
skill-project.toml
3

Add a skill

Pick the source that matches your workflow:
4

Install & index

5

Search & verify

You should see JSON results for installed skills. If empty, run fastskill install, then fastskill reindex, and search again.
6

Optional: serve UI

Browse http://localhost:8080/ for the bundled UI.The HTTP API is available under /api/v1/…, e.g.:
  • GET http://localhost:8080/api/v1/skills — list installed skills
  • GET http://localhost:8080/healthz — liveness probe
  • GET http://localhost:8080/readyz — readiness probe

Use-case quick recipes

  • Install skills from different sources
    • Public registry ID: fastskill add pptx@1.0.0
    • Local folder (editable): fastskill add ./skills/my-skill -e --group dev
    • Git repo with branch: fastskill add https://github.com/org/skill.git --branch main
    • Private Git marketplace:
  • Add standard-compliant skills (SKILL.md only)
    FastSkill will:
    • Extract the skill ID from SKILL.md (name or metadata.id)
    • Extract version from SKILL.md (metadata.version or version, defaulting to 1.0.0)
    • Display a warning suggesting fastskill init to add skill-project.toml

Troubleshooting

  • No skills found: ensure skill-project.toml exists at project root and rerun fastskill install.
  • Search returns nothing: run fastskill reindex after adding or installing skills.
  • Binary not on PATH: reinstall or move the binary to a directory in PATH (e.g., ~/.local/bin).