Drive Recommendations

When to use

Activate this skill when the user asks anything along the lines of:

• "Any advice on how I drive?"
• "How can I get more range?"
• "Give me recommendations on my drives."
• "What should I change in my driving habits?"
• "Tell me one thing to improve in how I drive."

If the user wants a full breakdown (median Wh/km, outliers list, baseline analysis) instead of focused advice, prefer the drive-efficiency-coach skill.

Prerequisites

You need a TeslaMate car_id. If unknown:

1. Call teslamate_get_cars to list available cars.
2. If exactly one is returned, use it silently.
3. If multiple, ask the user which one before continuing.

Workflow

1. Pull the dataset (current + previous period)

Call teslamate_get_car_drives twice:

• Current period — default: 30 days ago → now (RFC3339). If the user

specifies a different period, use that instead.

• Previous period — same duration, immediately preceding the current

window. For a 30-day current window, that means days -60 → -30.

Both calls use:

• car_id: target car
• min_distance: 5 (filters out parking-lot moves where Wh/km is meaningless)
• fetch_all: True whenever the window is longer than ~1 month — the

TeslaMate API pages at 100 entries.

Always pass both start_date and end_date. Skip the previous-period call only if the user explicitly says "just this month" / "no comparison".

2. Compute per-drive signals

For each drive:

• Distance in km (convert from miles if TeslaMate is configured that way).
• Energy used, from whichever field is present (consumption_kwh,

range delta × nominal Wh/km, or SoC delta × usable pack capacity).

• wh_per_km = energy_used_wh / distance_km
• avg_speed_kmh = distance_km / (duration_min / 60)
• Speed band: city (< 50 km/h), mixed (50–90), highway (> 90).
• Trip length: short (< 15 km), medium (15–60), long (> 60).
• Outside temperature if exposed.
• Skip drives where distance_km < 5 or energy_used_wh <= 0.

3. Compute the signals that matter

Compute the same signals on both periods so you can phrase recos as deltas:

• Median Wh/km (the user's baseline).
• Wh/km by speed band — flag any band that is ≥ 15% above the baseline.
• Wh/km by trip length — flag short trips if they consume ≥ 20% more

than medium trips (climate warm-up not amortized).

• Wh/km by temperature bucket if data is rich enough — flag cold drives

(< 5°C) that exceed the baseline by ≥ 20%.

• Highest single-drive Wh/km on ≥ 20 km — useful as a concrete example.

For each signal, compute the delta vs previous period (absolute Wh/km and %). Flag a habit as "improving" / "drifting" / "stable" based on the direction. Stable = delta within ±3% on Wh/km (driving is noisier than charging — keep the band tight to avoid false positives).

Be wary of confounders before phrasing a delta as a habit change: temperature swings (winter vs summer) or a long road-trip mixed into one period can move Wh/km by 20–30% on their own. Mention the confounder explicitly if you can see it in the data.

4. Pick 1–3 recommendations

Surface only the most impactful ones, anchored in the user's own numbers. Prefer signals that drifted in the wrong direction vs the previous period — those are the strongest hooks because they show a habit slipping. Examples — adapt the wording and the number to the actual data:

• "Highway Wh/km drifted from X to Y vs the previous 30 days (+Z%).

Cruising at 100–110 km/h instead of 120+ would save a meaningful chunk on long trips."

• "Median Wh/km went from X to Y. The biggest contributor is your

highway band — see above."

• "Short-trip Wh/km is X% above your medium-trip baseline (similar to

last period). Combine errands when possible — cabin warm-up isn't amortized on short hops."

• "Cold-weather drives (< 5°C) cost you ~X% more Wh/km this period —

precondition while plugged in to recover most of that overhead."

• "Your worst drive of the period was X Wh/km on Y km, dragging the

average up by Z%. One outlier drive can dominate a short window."

Also call out one clear improvement when you see it — positive reinforcement keeps the user engaged: "Median Wh/km dropped from X to Y — keep it up."

Avoid generic platitudes. Every recommendation must reference a number derived from this user's data.

5. Output format

1. Markdown summary (short, scannable):

• One-line context: current period, number of drives, total km,

median Wh/km, and the headline delta vs previous period (e.g. "median 168 → 182 Wh/km, +8%").

• The 1–3 recommendations as a bulleted list, each one with:
• a short headline (the change to make),
• the data point that triggered it, **with the previous-period

value for comparison whenever available**,

• the expected benefit (% Wh/km, km of range, or kWh saved).
• Optional one-line positive note if a habit visibly improved.

1. Then render the recommendations as a Prefab UI Dashboard.

• First call generative_search_prefab_components to discover the

components currently available — newer / richer ones may exist.

• Then call generative_generate_prefab_ui with a Dashboard

layout (outermost) wrapping the components that best illustrate the recommendations. Better too many components than too few.

Lean into whatever discovery returned, prioritising:

• Metric cards for the key numbers anchoring each tip (with the

previous-period number as a delta where you have it).

• Progress for ratios (e.g. share of highway km).
• LineChart / AreaChart for the Wh/km trend by drive, day or

week.

• RadarChart for multi-axis comparisons (efficiency vs speed band

vs trip length vs temperature, normalised).

• PieChart for breakdowns (city / mixed / highway, short / medium /

long).

• DataTable for the outlier drives worth inspecting.
• Badge for verdicts ("improving", "drifting", "stable") and

Label for compact captions.

• show_map to map outlier drives' start/end on a map — addresses

are in start_address / end_address.

Follow the global Prefab UI conventions from the MCP server instructions: Dashboard outermost, discovery-first, favourite set when several options fit equally.

Pitfalls

• Short drives mislead. Drives < 5 km are dominated by climate/start-up

energy. Always filter them out before computing the baseline; they're only useful as the basis for the "combine short trips" recommendation.

• Energy data is sometimes missing. TeslaMate has gaps. Skip drives

with missing energy rather than imputing zeros.

• Units vary per TeslaMate install. Default to km, but if distances

look suspicious (e.g. all values < 5 for normal commutes), inspect a sample drive's units field and convert.

• Small samples kill confidence. Under 10 valid drives, lead with a

single recommendation and say so explicitly — don't stretch noisy data into three insights.

• Don't blame habits for temperature. A 20% Wh/km jump between two

30-day windows that straddle the seasonal shift is mostly weather, not the user's foot. If the temperature buckets shifted between periods, say so before phrasing the delta as a recommendation.

• Don't double up with `drive-efficiency-coach`. This skill is

recommendations only; if the user asked for a full breakdown, hand off to that skill.