Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
npx skills add https://github.com/Tzeusy/butlers --skill energy
La commande reste sur une seule ligne. Faites défiler horizontalement pour la vérifier avant de la copier.
Vous préférez une copie locale ? Téléchargez les fichiers actuellement disponibles dans SkillsMP.
Analyze home energy consumption, detect anomalies, and recommend optimizations.
Skill: Energy Awareness
Purpose
Track and analyze home energy consumption to help user understand where energy goes, identify
inefficiencies, and make optimization decisions. Generate weekly energy digests with trends,
device rankings, and actionable recommendations. Alert on anomalies (unexpected spikes, always-on devices).
When to Use
Use this skill when:
User asks about energy consumption ("What's my energy usage?", "Which devices use the most?")
System needs to analyze energy trends for weekly digest
Anomalous consumption is detected (unusual spike, device left on unexpectedly)
User wants optimization recommendations
Workflow
Step 1: Retrieve Energy Data
To answer any energy question:
Determine the time period from user's request or use defaults:
"Today" → current day
"This week" → last 7 days
"This month" → last 30 days
No context → use "this week" as default
Call energy_get_consumption() with:
period: Time period (today, this_week, this_month, custom date range)
granularity: Optional — hourly, daily, or summary (default: summary)
Call energy_get_devices_by_usage() to rank top consumers:
period: Same as above
limit: 5-10 devices (show top consumers)
Call energy_get_peak_times() to identify demand patterns:
period: Same as above
Step 2: Analyze Against Baselines
Compare current consumption to historical patterns:
Call memory_recall() to retrieve stored energy baselines:
Look for energy_baseline facts about typical consumption by device or time
Calculate deviations:
Is this week higher/lower than typical?
Are peak times consistent with known patterns?
Are any devices consuming more than usual?
Identify anomalies:
Device always-on when it should be off (e.g., AC running in winter)
Unexpected spike at unusual time
Device consumption 20%+ above baseline
Step 3: Store Energy Patterns as Memory
After analyzing consumption:
Call memory_store_fact() to persist key findings:
subject: Device name or "energy"
predicate: energy_baseline, energy_spike, or energy_pattern
content: Concise description of pattern or anomaly
permanence: standard (patterns may shift seasonally)
importance: 5-7 (depends on significance)
tags: Device names, time periods, severity (e.g., ["hvac", "baseline", "winter"])
Example facts:
"HVAC typically uses 40% of daily energy in winter, peaks 7-9am and 6-8pm"
"Water heater ran for 6 hours yesterday instead of usual 2 hours — possible malfunction"
"Dishwasher left on standby uses 5W constantly — minor waste"
Step 4: Respond to Energy Queries
When user asks about consumption (e.g., "What's my energy usage today?"):
Retrieve data (Step 1)
Analyze patterns (Step 2)
Compose response via notify() (answer mode):
Current usage: "Today: 18 kWh" or "This week: 120 kWh"
Top consumers: "Top 3: HVAC (40%), Water heater (25%), Kitchen appliances (15%)"
Trend: "8% below your weekly average" or "Up 12% due to unusually cold weather"
Key insight: Highlight any anomalies or seasonal patterns
Optional suggestion: "Want me to suggest ways to reduce water heater usage?"
Step 5: Generate Weekly Energy Digest (Scheduled Task)
Run on schedule (e.g., Sunday 9am):
Retrieve last 7 days of data via energy_get_consumption(period="this_week")
Retrieve peak times and device rankings
Analyze trends:
Is consumption trending up or down?
Are peak times consistent?
Any anomalies this week?
Compared to previous week?
Calculate efficiency metrics:
Call energy_get_efficiency_metrics() for year-over-year or month-over-month comparison
Estimate cost savings achieved vs. baseline
Generate recommendations:
Call energy_suggest_optimization() for top energy consumers
Expected savings: "Estimated savings: 5-10% of HVAC usage ($3-6/month)"
Call-to-action: "Should I adjust thermostat schedules?"
Get user consent before implementing changes
Key Behaviors
Energy Queries Are Conversational
One metric per user message:
Good: "What's my energy usage today?"
Avoid: "What's my usage today, peak times, and top devices?" (ask one, offer others in response)
Provide Context, Not Just Numbers
Never just say "18 kWh". Always provide context:
vs. typical usage ("8% below average")
vs. time ("lower than last week because weather was milder")
vs. goal ("on track for monthly budget")
Anomaly Severity Levels
Low (include in digest): Standby power (TV, chargers using 1-2W)
Medium (proactive alert): Device consuming 10-20% more than baseline
High (immediate alert): Device consuming 2x baseline, always-on when should be off, or safety concern
Avoid Overwhelming Recommendations
Give 2-3 suggestions max, ranked by impact. Offer to dive deeper if user wants.
Connect to Comfort
Remember comfort preferences when suggesting optimizations:
Don't suggest "lower heating to 62°F" if user prefers 70°F
Suggest alternatives: "Running AC at night uses 30% less energy due to cooler outside temps. Want to try sleeping cooler?"
Multi-Turn Example: Energy Optimization (HA Tools)
This example uses the native Home Assistant tools. It illustrates a realistic 3-turn conversation
where a user investigates high HVAC usage and adjusts settings.
Turn 1 — User: "Show me my top energy consumers"
Actions:
ha_list_entities(domain="sensor") — find energy monitoring sensors (look for energy, kwh, power in entity IDs)
notify(channel="telegram", message="HVAC analysis:\n• Running 14h/day on average (high for this time of year)\n• Peak usage: 7-9am and 6-9pm\n• 18% above your usual baseline\n\nSuggestions:\n1. Lower nighttime setpoint by 2°F (saves ~10% on heating)\n2. Pre-cool/pre-heat 30min before peak demand, then let it coast\n3. Check/replace air filter — dirty filters increase runtime\n\nWant me to adjust the night setpoint?", intent="reply", request_context=...)
Turn 3 — User: "Set bedroom temperature 2 degrees lower at night"
Actions:
ha_get_entity_state(entity_id="climate.bedroom") — get current setpoint (e.g., 70°F)
memory_store_fact(subject="bedroom", predicate="comfort_preference", content="user prefers bedroom 2°F cooler at night; setpoint reduced to 68°F to save HVAC energy", permanence="stable", importance=7.0, tags=["temperature", "comfort", "bedroom", "night", "energy"])
notify(channel="telegram", message="Bedroom night temperature set to 68°F (was 70°F). This should trim HVAC runtime and reduce your monthly usage.", intent="reply", request_context=...)
Exit Criteria
energy_get_consumption() was called to retrieve usage data
energy_get_devices_by_usage() and/or energy_get_peak_times() were called
memory_store_fact() was called to persist baseline, pattern, or anomaly
User has been notified via notify() with:
Current consumption OR
Weekly digest OR
Anomaly alert OR
Optimization recommendations
Session exits without starting new workflow
Common Failure Modes and Recovery
User Asks About Future Optimization ("Will this save money?")
Cannot predict precise savings without knowing actual changes
Provide estimate based on baseline: "Lowering AC 2°F typically saves 3-5%. For you, that's roughly $2-4/month"
Offer to track actual savings after change is implemented
Energy Data Has Gaps
Alert user: "Energy meter was offline from 2-4pm. Digest shows partial data for today."
