skills-vis 0.1.3

Monitor, manage, and visualize AI agent skills, execution flow, and skill folders in one unified observability platform.

Skills Vis

An interactive browser for your AI agent skill libraries — visualize your skills as a graph, edit them in place, and track which ones your agent uses in real time.

What it does

• Browse skills as a graph — navigate your skills folder as a left-to-right tree; expand, collapse, and select nodes with a click.
• Edit skills in the browser — click any file to open a syntax-highlighted editor with live markdown preview.
• Manage skills — create, rename, and delete files and folders directly from the UI.
• Import skills from GitHub / GitLab / local paths — discover and pull in skills from remote repositories with conflict detection.
• Live usage monitor — see which skills your AI agent invokes highlighted on the graph in real time; replay past sessions.
• MCP server — let Claude Code manage recording sessions directly via MCP tools. (Coming soon)

Installation

pip install skills-vis

Requires Python 3.10+.

Quick start

skills-vis

Open http://127.0.0.1:8001 in your browser, paste the path to your skills folder, and click Load.

Live agent monitoring

To see skill invocations highlighted on the graph as your agent works:

1. Click ⏺ Record in the toolbar to start a session.
2. Load the skills-monitor/SKILL.md skill into your agent session.
3. Work normally — every skill the agent invokes will glow green on the graph for a few seconds.
4. Click Recording… to stop. The session is saved and replayable from the Skills drawer.

Importing skills from a remote source

Open the Skills drawer (☰ Skills) and click Import Skills. Paste a GitHub URL, GitLab URL, or local path. Skills Vis will scan for SKILL.md files, let you preview them, and copy the ones you select into your active folder.

Development

See dev/ for setup, running locally, tests, and build instructions.

License

MIT