Vironnimo

home-assistant

Inspect and safely change advanced Home Assistant configuration through the bundled WebSocket script, especially Lovelace dashboards, dashboard metadata, and registry-backed planning. Use when the user asks to create, redesign, back up, export, restore, or deeply edit Home Assistant dashboards or needs Home Assistant configuration beyond entity state inspection and service calls. Do not use for ordinary device control when the ha_* Tools suffice.

browser-use

Drive an isolated browser through the bundled Python wrapper around agent-browser. Use when a task requires interactive navigation, accessibility snapshots, clicking, filling forms, keyboard input, scrolling, or visual page inspection that web search and web fetch cannot complete.

computer-use

Inspect and operate desktop application windows through the bundled Python wrapper around cua-driver. Use when the user explicitly asks the Agent to interact with a graphical application and ordinary vBot Tools cannot perform the task. Prefer browser-use for web pages and ordinary Tools for files, processes, and configuration.

coding-agents

Operate Codex, Claude Code, OpenCode, and other interactive coding-agent CLIs as persistent shared TUI sessions through terminal, including explicit model, reasoning or effort, Agent, profile, permission, sandbox, and other startup settings. Use when delegating coding work, selecting how a coding agent should start, supervising it, answering its questions or approvals, checking progress, or continuing an existing coding-agent session. Do not use to operate vBot itself; that is the vbot-cli skill.

teach

Teach the user a topic, concept, or skill so they actually understand it and can do it themselves — not just hand over information. Use when the user asks to be taught, wants to learn or understand something, says they don't get a concept, asks you to explain so it sticks, or wants to be walked through how something works. Examples — "teach me X", "explain X so I get it", "help me learn Y", "I don't understand Z", "walk me through how W works", "I have no idea about V". Do not use for quick factual lookups where the user just wants the answer and moves on.

vbot-cli

Configure, inspect, and operate vBot through the vbot CLI, including locating the vBot root and runtime data. Use when asked to start, stop, restart, update, or uninstall vBot, set up provider credentials (API key or OAuth), find vBot files or diagnose stored state, list/add/edit/remove agents, projects, sessions, channels (Telegram, Discord), cron jobs, Bootstrap Runs, prompts, skills, pinned Memory entries, extensions, or settings, inspect complete Model data, or configure specialized Task Models such as TTS/STT including model-specific voices and options — as well as to inspect tools, logs, debug traces, Provider subscription usage and limit history, and Session usage statistics (tokens, runs, errors, tool and skill usage).

weather

Get current weather conditions and multi-day forecasts for any city, region, airport code, landmark, or coordinates via the free wttr.in service (no API key), using the web_fetch tool. Use when the user asks what the weather is like, whether it will rain, how hot or cold it is, or wants a forecast for travel planning. Examples — "what's the weather in Berlin", "will it rain in Paris tomorrow", "how hot is it in Tokyo right now", "weather forecast for my trip to London".

pdf

Create, restyle, inspect, and visually verify polished PDF documents. Use when the user asks to create a PDF, turn supplied content into a PDF, improve an existing PDF, review PDF layout, or deliver a professional report, brief, handout, or printable document as a .pdf file.

free-models

Conserve budget by delegating suitable Sub-Agent work to free-of-charge Models (OpenRouter free variants, OpenCode Zen free tier) and by helping the user set up a free-capable Provider when none is configured. Use when the user asks to save costs or mentions free models, a low budget, or spending nothing, when many small independent subtasks can run on weaker Models, or before fanning out bulk Sub-Agent work. Do not use for critical, security-sensitive, or irreversible work that needs top-tier quality, and never as a license to downgrade the main conversation's own Model.