Add a Skill

Import existing skills onto your canvas from a public GitHub repo or your own machine - no CLI, no zip files.

You don't have to build every skill from scratch. Add a skill brings existing skills straight onto your canvas - paste a public GitHub link or drop a SKILL.md from your computer, preview what's inside, and pick what to import.

Each skill you import becomes a skill piece on your canvas, with its full SKILL.md and any bundled files preserved so you can read, edit, download, or share it.

Two kinds of import. This page covers the visual import that adds skills to your canvas (a pie). If you instead want to manage skill files on your machine and link them into Claude Code, Codex, and Cursor, use the ppie CLI.

Ways to start

There are three ways to open the importer:

On an empty canvas, click Add a skill.
From the header, click + Add Skill at any time.
Drag and drop a SKILL.md file or a skill folder anywhere on the canvas.

Empty canvas with Create a prompt and Add a skill buttons — Start a pie with Create a prompt or Add a skill. You can also drop a SKILL.md or folder anywhere to import.

Importing a skill

Choose a source

Paste a public GitHub URL, pick one of the Popular Skills, or drop a local file.

Import Skills modal with a GitHub URL field, Popular Skills, and a local drop zone — Paste a GitHub URL, choose a popular source, or drop a local SKILL.md or folder.

GitHub - point at a repo, a folder, or a single SKILL.md:

Source	Example
Repo	`https://github.com/owner/skills`
Folder	`https://github.com/owner/skills/tree/main/skills/pptx`
Single skill	`https://github.com/owner/skills/blob/main/skills/qa/SKILL.md`

Popular Skills - curated, ready-to-import sources so the importer is never blank:

Anthropic - Official Claude Skills
OpenAI - Official Codex Skills
Matt Pocock - Skills for real engineers

Local - drop or Browse a SKILL.md, a skill folder, or a .zip / .tar / .tgz archive. Local files are previewed in your browser before anything is added.

What isn't supported. Private repos, GitHub profile URLs (like github.com/owner), and Gists are rejected before preview. Prompt Pie reads public GitHub content only and never uses your GitHub credentials.

Preview and select

Prompt Pie scans the source and lists every SKILL.md it found. Nothing is added to your canvas yet.

Preview list showing 29 skills found with checkboxes, search, and Select all — Preview every discovered skill, then pick which ones to import.

Each row shows the skill's name, path, size, and a short description. From here you can:

Select the skills you want, or use Select all.
Search to filter a long list - your selection is kept even when rows are filtered out.
Expand the +N files badge to see a skill's bundled files (references, scripts, assets) before importing.
Spot a large badge on heavier skills.

For big, deeply-nested skills, the preview splits the bundle into To import (the SKILL.md plus the files it references) and Not imported (a compact folder tree that links back to GitHub). You always import the skill's real working set without dragging in megabytes of generated files.

You can import up to 50 skills at once. A single skill is selected for you automatically.

Import to your canvas

Click Import Skills and each selected skill drops onto the canvas as its own skill piece, laid out in a tidy grid and added as a single undoable step.

Every imported piece keeps the raw SKILL.md as its source of truth, so you can edit it on the canvas and still download, copy, or share it as a spec-compliant skill. A footer notes where it was imported from.

The Skill Inspector

Expand an imported skill to open the Skill Inspector - a read view for everything that came with it.

Skill Inspector showing a rendered SKILL.md, a bundle and source sidebar, and the skill's file structure — The Skill Inspector: rendered SKILL.md, the files that came with the skill, and where it came from.

Rendered SKILL.md on the left, with the bundled files listed alongside it.
Files, grouped by role: Bundle (with an "X of N stored" count), References, Scripts, Assets, and Other. Relative links in the skill navigate between bundled files.
Source - the origin repo, ref, and path, with Open on GitHub.
Metadata parsed from the skill's frontmatter.

Scripts are stored inert. Prompt Pie never runs, installs, or follows commands from an imported skill. Script files are kept as text so you can read them - nothing more.

Is it safe?

Importing a skill only ever reads and displays content. Specifically, Prompt Pie:

Reads public GitHub content only - no private repos, OAuth, or access tokens.
Never executes scripts or install/setup commands from imported skills.
Previews local files in your browser before anything is added.
Fetches GitHub content server-side with size caps, timeouts, and rate limits.
Scans imported text for risky patterns - secrets, hidden or bidirectional characters, and shell or package-install commands - and surfaces them as warnings.

Visual import vs. the CLI

Both bring skills in, but they land in different places:

	Add a skill (this page)	`ppie` CLI
Imports into	Your canvas (a pie)	Your local filesystem (`~/.promptpie/skills/`)
Best for	Browsing, organizing, and sharing skills visually	Linking one source-of-truth skill into Claude Code, Codex, and Cursor
Runs where	In the app	In your terminal

Want the underlying format? See Agent Skills for the SKILL.md spec that every imported and exported skill follows.

On this page