skills-vis 0.1.5

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.

User Guide (English) | (中文)

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

Click the Browse… button (or the import icon in the toolbar). Paste a GitHub URL, GitLab URL, or local path — for example https://github.com/mattpocock/skills. Skills Vis will scan for SKILL.md files, let you preview and select them, then copy the ones you choose into your active folder.

Development

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

License

MIT