SkillScope - See Exactly What a Claude Code Skill Does Before You Run It
You’ve installed a skill from GitHub. It misfired. Now you’re staring at 300 lines of markdown trying to figure out what it was supposed to do.
SkillScope fixes this. One SKILL.md in, one interactive flowchart out.
What It Shows
SkillScope parses every section of a skill file and renders it as a visual flow:
- Triggers (cyan) - what activates the skill
- Steps (gray) - the execution sequence
- Tool calls (lime) - which Claude tools get used
- Branches (amber) - conditional decision points
- Outputs (rose) - what gets returned
- Chains (violet) - when the skill hands off to another skill
Reading a skill diagram takes 10 seconds. Reading the raw markdown takes 5 minutes.
The Skill Quality Score
Every parsed skill gets a 0–100 score across 5 categories:
| Category | Points | What it checks |
|---|---|---|
| Frontmatter | 20 | name, description, allowed-tools |
| Structure | 25 | sections, When to Use, Steps, Output Format |
| Clarity | 20 | numbered steps, tool references, code blocks |
| Logic | 15 | branching, error handling, chain awareness |
| Best Practices | 20 | length, templates, focus, detailed descriptions |
Each issue flags a severity level and a specific fix recommendation. Useful for auditing skills you’re about to install, and even more useful for reviewing skills you wrote yourself six weeks ago.
Zero Setup
Single HTML file. Opens directly in any browser.
git clone https://github.com/silvesterdivas/skillscope.git
open app.htmlPaste a SKILL.md, browse to a file, or pick one of the built-in examples. No build step, no dependencies, no config.
Why I Built It
I was using 15+ skills across my Claude Code setup and had no reliable mental model of what each one actually did. When something went wrong, debugging meant re-reading files I’d already read. SkillScope makes the visual model persistent - you look at the flowchart once and remember it.