Subagents and skills
Follow a Claude's own agents as they work, and give an agent skills of its own.
Subagents in the sidebar
A Claude can run agents of its own, defined in its project's .claude/agents. They run inside that one Claude tab, with no extra terminals. While they work, vyberun gives each a line under the Claude tab in the sidebar: what it's doing now, and for how long. A finished one says done for a few minutes afterwards.
Following one
Click a subagent's line to follow its work beside the terminal: what it was asked, the files it read and changed, the commands it ran and what came back. The panel is read only: nothing is sent to the agent. Esc or ✕ closes it.
New and stuck
- new: it finished while you were elsewhere. If vyberun isn't in front and notifications for long commands are on, Windows tells you too.
- stuck? in amber: it keeps hitting the same wall, such as commands blocked by a hook, the same command again and again, errors in a row, or a long silence. Its panel says why.
Skills
A saved agent can bring skills: instructions Claude Code loads for that tab alone.
- Open Settings ▸ Agents and edit the saved agent.
- Add .md or .zip files…: a
.mdis a skill itself, a.zipis a skill's folder. - Start the agent on a card (⋯ ▸ Start agent). Its skills are there in that tab.
Edit a skill's file any time, in any editor: an agent started afterwards reads the new version.
vyberun