| name | weather |
| description | Fetch current weather via Open-Meteo using the user's location from user_profile.json. Always deliver warm, contextual, actionable advice (never raw numbers). |
Weather Skill
When to use this skill
Any time the user asks about the weather, temperature, forecast, “is it nice out?”, what to wear, outdoor plans, etc.
Core workflow (always follow exactly)
-
Read location
Use read_file to load ~/.katana/defaults/user_profile.json'(check project root first, then /.config/kilo/user_profile.jsonor/user_profile.json`).
Expected structure:
"location": {
"latitude":,
"longitude":,
"city":,
"state":,
"country":,
"timezone":
},
```
Fetch data (no API key needed)
Run this bash command (replace ${LAT} and ${LON}):Bashcurl -s "https://api.open-meteo.com/v1/forecast?latitude=${LAT}&longitude=${LON}¤t_weather=true&daily=weathercode,temperature_2m_max,temperature_2m_min,precipitation_probability_max&timezone=auto"
Interpret & contextualize
Turn raw temp + weathercode into a human-friendly summary.
Never just say the temperature.
Always add real-life impact and friendly advice.
Exact tone & style you must match
"72 and clear — good day to work outside if you need a reset."
"Rain all afternoon — cancel that outdoor lunch plan."
"61° and overcast, light wind. Comfortable for a run but you’ll want a light layer."
Bonus examples
“48° and pouring in Sandusky — definitely cancel that outdoor lunch plan and stay in to code.”
“Clear skies, 72°F, light breeze — perfect reset day. Go touch grass after you ship that PR.”
“High of 55° with 80% chance of afternoon showers — morning is dry if you want to grab coffee outside.”
Pro tips for the agent
Mention city name when relevant.
Include short-term outlook (afternoon/evening) if the user might care.
Keep it warm, concise, and useful — like a helpful local friend.
