Skill Editor

A full-featured editor for writing, reviewing, and managing your AI skills - right in the browser.

Monaco Editor

The skill editor is powered by Monaco - the same engine behind VS Code. It provides a familiar editing experience with features built for SKILL.md files:

Syntax highlighting - Markdown body and YAML frontmatter are highlighted with distinct colors so you can tell metadata from instructions at a glance.
Word wrap - long lines wrap automatically. Toggle word wrap from the editor toolbar.
YAML validation - the editor validates your frontmatter in real time. Missing required fields, invalid types, and malformed YAML are flagged with inline error markers before you save.
Keyboard shortcuts - standard Monaco shortcuts work out of the box: Cmd/Ctrl+S to save, Cmd/Ctrl+Z to undo, and Cmd/Ctrl+Shift+P for the command palette.

Reference Files

Skills often need supporting material - templates, example configs, helper scripts. The editor lets you manage these directly alongside the skill.

Create inline- click "New file" in the reference panel to add a new file. Give it a name and start typing. The file is saved alongside the skill and synced to all targets.
Edit in-place - click any reference file to open it in a secondary editor tab. Changes are saved together with the skill.
Folder organization - reference files are organized into conventional directories: references/ for templates and examples, scripts/ for executable helpers, and assets/ for images or diagrams. You can also create custom directories.

Learn more about structuring reference files in the Writing Effective Skills guide.

AI Review

Click "Review" in the editor toolbar to get an AI-powered quality assessment of your skill. The review scores five dimensions:

Dimension	What It Measures
Clarity	Are instructions unambiguous and easy to follow?
Specificity	Does the skill use concrete details instead of vague language?
Structure	Is it well-organized with headings, steps, and logical flow?
Completeness	Does it cover edge cases, errors, and output format?
Actionability	Can an AI tool execute it without follow-up questions?

Each dimension is scored 1 to 5. The review also provides specific suggestions and an "Apply fix" button for each finding. Clicking "Apply fix" inserts the suggested improvement directly into the editor so you can review and accept it.

Version History

Every save creates a new version automatically - no manual versioning needed. The version panel shows a chronological list of all previous versions with timestamps.

Side-by-side diffs - select any two versions to see a highlighted diff. Additions, deletions, and modifications are shown in the familiar green/red format.
Rollback- click "Restore" on any previous version to revert the skill to that state. A new version is created for the rollback so the full history is preserved.
Automatic versioning - the version number in the YAML frontmatter is incremented automatically on each save. You can also set it manually if you prefer semantic versioning.

Security Scanning

Skills are scanned automatically for security issues before every save. The scanner checks for:

Hardcoded credentials - API keys, tokens, passwords, and secrets embedded in the skill body or reference files.
Prompt injection patterns - instructions that could cause an AI tool to bypass safety guidelines or execute unintended actions.
Unsafe patterns - commands that delete files, modify system configurations, or access sensitive directories without appropriate guards.

If a security issue is found, the save is blocked and you are shown a detailed explanation with a suggested fix. You can override the warning if the flagged content is intentional.