[rishsriv]

The FactIQ skill now tells the model about two new satellite datasets: wind_power and humidity_power. With them it can check whether a change in air pollution came from emissions or from the weather, and which way smoke was blowing. The datasets themselves are served by the FactIQ API. This change updates the documentation so the model knows they exist and how to read them.

The model needs to know how to read wind direction

wind_power returns wind_from_deg, the compass bearing the wind blows from. The tool works it out from the average wind over a region and a period, so winds from opposite directions cancel each other. The satellite reference now says to read wind_steadiness first. A value near 1 means the wind kept one direction. A value near 0 means the direction says little. The reference also says to use a state or a box, not a whole country, when saying where smoke came from.

What changed

• references/data/satellite.md: two new rows in the datasets table, and a note on reading wind direction and steadiness.
• skills/factiq/SKILL.md: the get_geo_data row lists the two datasets.
• evals/mocks/factiq/_tools.json: the mock tool description matches the live tool text.
• Version 0.41.3 → 0.42.0 in both plugin manifests.

Example

Before this change, the skill did not list wind data. A question such as "was Haryana's November 2025 smoke blown in from the north-west?" had no data to answer it.

After: get_geo_data("wind_power", "India/Haryana", "2025-10-01", "2025-11-30") returns

columns: month, wind_speed_ms, wind_speed_max_ms, u_wind_ms, v_wind_ms, wind_from_deg, wind_steadiness
["2025-10", 1.86, 2.88, 0.07, -0.01, 278, 0.04]
["2025-11", 2.43, 3.79, 1.05, -1.14, 317, 0.64]

The skill tells the model to ignore October's direction (steadiness 0.04). November's wind blew steadily from the north-west (317°).

Tests: pytest tests passed (80 passed).

🤖 Generated with Claude Code