Audit whether the organisation's AI spend actually paid — measured against baselines, not vendor math or vibes. Use when a CFO asks what the AI tools returne...
安全
appliance-energy-audit
试用Use when you want to know which appliances actually drive your electric bill, whether a bill spike is explained by your usage, if standby/vampire draw is worth addressing, whether replacing an old fridge/dryer/AC pays back, or when modeling tiered utility rates — builds a ranked cost-per-appliance table from watts, duty cycle, and usage hours, reconciles it against your real bill, and computes replacement payback.
它能做什么
Use when you want to know which appliances actually drive your electric bill, whether a bill spike is explained by your usage, if standby/vampire draw is worth addressing, whether replacing an old fridge/dryer/AC pays back, or when modeling tiered utility rates — builds a ranked cost-per-appliance table from watts, duty cycle, and usage hours, reconciles it against your real bill, and computes replacement payback.
技能文档
Appliance Energy Audit
Overview
Your utility bill says you used 850 kWh last month — but says nothing about which device used them. This skill turns a walk-through inventory of your home ("fridge from 2012, gaming PC, window AC…") into a ranked cost table: what each appliance consumes per month and year, what share of the bill it represents, how much pure standby ("vampire") draw is costing, and — when you feed in your actual bill kWh — whether the model matches reality or something is running more than you think.
It also answers the eternal appliance-store question: is the new efficient model actually worth it? The replace command computes monthly/yearly savings and payback months for any old-vs-new pair at your electricity rate.
When to Use
- "Which appliances are costing me the most?" / "Why is my electric bill so high?"
- "Is it worth replacing my 15-year-old fridge / electric dryer / AC?"
- "How much do all these always-on devices (router, console, smart speakers) cost me?"
- "My bill jumped — does my usage actually explain it?"
- "I have tiered rates — how does that change the picture?"
- Don't use for: gas appliance costs, whole-house heat-loss modeling, or real-time monitoring (this is an analytical model, not a meter).
The Model
Each appliance is characterized by four numbers (all have sensible library defaults):
| Field | Meaning | Example |
|---|---|---|
watts | draw when active | 150 W fridge compressor |
duty | fraction of active time it actually draws (compressors cycle) | 0.35 |
hours_day | hours/day it's on | 24 for a fridge, 0.5 for a dryer |
standby_w | idle draw the other hours | 3 W for a TV |
monthly_kWh = (watts × hours_day × 30 × duty + standby_w × (24 − hours_day) × 30) × qty ÷ 1000
cost = kWh × rate (flat)
= tiered_cost(kWh) (tiered rate plans)
Vampire draw is the standby term summed across appliances. Full derivation and the tiered-rate marginal-cost approximation are in references/energy-model.md.
Commands
# Browse the 44-appliance preset library (watts, duty, default hours, standby)
python3 scripts/energy_audit.py library
# Quick single-appliance estimate
python3 scripts/energy_audit.py estimate "my dryer,electric-dryer,0.5" --rate 0.25
# Full audit — repeat -a for each appliance, ranked by monthly cost
python3 scripts/energy_audit.py audit -a "fridge,fridge" -a "tv,tv-oled-55,5" \
-a "bulbs,led-bulb,4,8" -a "gaming pc,gaming-pc,3" --rate 0.17
# From a JSON inventory file, with tiered rates and bill reconciliation
python3 scripts/energy_audit.py audit -f home.json --tiers "0.12:500,0.15:1000,0.20:" --calibrate-to 850
# Model vs actual bill only
python3 scripts/energy_audit.py calibrate -f home.json --bill-kwh 620
# Replacement math: old fridge vs $800 efficient model
python3 scripts/energy_audit.py replace --old "old fridge,fridge" \
--new "efficient fridge,120,0.3,24" --price 800 --rate 0.17
# See everything at once on sample data
python3 scripts/energy_audit.py example
Appliance shorthand: name[,preset|watts][,hours_day][,qty][,standby_w] — with a preset, numbers mean hours/qty/standby; without one, the first number is watts. JSON spec files are lists of the same fields in long form ({"name": ..., "preset": ...} or {"name": ..., "watts": ...}).
Workflow
- Inventory your home: walk room to room; for each device run
libraryto find its preset or read the nameplate watts. - First pass:
auditwith defaults — the preset library already encodes realistic duty cycles. - Calibrate: get your bill's monthly kWh, run with
--calibrate-to. If the model says "undercounted", your heating/cooling or an aging compressor is running harder than rated — raise thosehours_day/dutyvalues and re-run. - Act: attack the top of the table. Use
replaceon any aging big-ticket row to get payback months; put high-standby devices (flagged automatically) on switched power strips. - Re-check quarterly or after any bill spike — the diff between runs is the story.
Common Pitfalls
- Forgetting duty cycle for compressor appliances. A fridge is "on" 24 h/day but its compressor runs ~35% of the time. Using duty 1.0 triples the estimate. The presets encode this — prefer them over nameplate watts.
- Using nameplate watts as typical draw. Ratings are peak/startup, not average. A desktop PSU rated 600 W averages ~200 W. Use measured or library values; nameplate only as a last resort.
- Ignoring tiered rates when you have them. On tiered plans the next kWh can cost 60% more than the first — flattening to one rate distorts which appliance matters. Pass
--tiers. - Trusting the model over the bill. If
calibrateshows a big gap, the model is wrong somewhere — not your meter. Find the missing/underestimated device (almost always HVAC run-time or water heating) before making purchase decisions. - Double-counting qty.
-a "bulbs,led-bulb,4,8"means 8 bulbs at 4 h/day. If you also list bulbs individually you'll count them twice.
Verification Checklist
-
python3 scripts/test_energy_audit.py→ ALL TESTS PASSED (63 assertions) -
python3 scripts/energy_audit.py libraryprints the preset table -
python3 scripts/energy_audit.py examplerenders a ranked audit + calibration + replace verdict -
audit --calibrate-togap is under ±10% before trusting per-appliance numbers
相关技能
Analyze electricity, water, and gas bills to detect anomalies, compare usage, forecast costs, and suggest savings.
Decide whether to fix, replace, or recycle a broken item. Takes item type, age, symptoms, and repair estimate, then produces a scored recommendation across cost, lifespan, sentimental value, and environmental impact.
Manage daily bills, expenses, and recurring payments with tracking, reminders, and analysis. Use when the user asks about bill management, expense tracking,...
Use when comparing apartments or rental homes and torn between options, when deciding if a listing is actually affordable, when touring rentals and needing to score them consistently instead of by vibes, when the advertised rent hides fees/parking/pet rent/utilities/commute, when preparing to negotiate rent or lease terms, or when defining what you actually need (bedrooms, pets, commute ceiling) before searching — applies hard constraints first (budget, commute, bedrooms, pets, move date) to kill unqualified listings, scores survivors on 16 weighted criteria YOU prioritize, computes the true all-in monthly cost with every fee amortized, runs affordability analysis (30/33/50% rules, 3x-income approval math), compares finalists side-by-side with the quality-vs-cost premium quantified, and generates a leverage-based negotiation script with anchoring targets.
Collect debt and mortgage inputs, call a plans API, and return payoff strategy comparisons (snowball, avalanche, refinance) with concise recommendations and...