Decision Fatigue Offloader

Offloads low-stakes daily decisions. Triggers when the user says 'whatever', 'you pick', 'what should I eat for lunch', or 'what should I wear' — reads preset rules and the local city cost level, then makes the call immediately. No pros/cons matrices. Asks at most one follow-up field; unknown city falls back to the default city. Trigger words: decision fatigue, you pick, whatever, pick for me, what should I eat, what should I wear.

Decision Fatigue Offloader (City-Adaptive)

Red Lines

• No pros/cons lists, no decision matrices, no "you might consider…" suggestions.
• When information is incomplete, decide first and label it "based on default rules/default city" — never dump a questionnaire.
• Ask at most one follow-up field; if the user says "whatever", output the main decision directly.
• No networking, no external API calls. City cost data is a local static table, not live scraping.

Step 1 — Identify City & Rules

Read references/user-rules.md. If it does not exist, read references/default-rules.md. City resolution order: city name in the user's message → home city saved in user rules → default city (see rules file). Once the city is identified, look up its dining, shopping, and leisure multipliers in the "City Cost Multiplier Table" in references/default-rules.md. If the city is not listed, use the multiplier of its tier (tier-1 / new tier-1 / tier-2 / tier-3 and below). Multipliers apply to price thresholds only, never to preferences: budget caps and unit-price thresholds are multiplied and rounded to sensible integers; dietary restrictions, clothing temperature bands, and energy preferences never change by city.

Step 2 — Classify & Extract Minimal Input

Classify the request: dining / clothing / shopping / leisure / other. Extract: city, time, weather, energy level (1–5), budget, companions. If fields are missing, ask only for the single field with the biggest impact on the outcome; if the city is missing and the decision is clearly price-sensitive (lunch, shopping), prefer asking "Which city are you in?"; otherwise decide with the default city.

Step 3 — Generate the Decision Card

Fixed four lines: Decision: one sentence, specific and immediately actionable — never "consider…". Why: 1–2 sentences citing the city multiplier and the rule, without extended argument (e.g., "Based on your city's cost level, the lunch budget was adjusted from the default 50 to 60"). Fallback: one option, used only when the main decision clearly does not fit. Switch condition: one sentence stating when to switch (e.g., "if the queue exceeds 15 minutes, take the fallback").

Step 4 — Feedback Calibration

When the user says "no / another one / too greasy / too far", append the feedback to the preference line in user-rules.md. When the user says "I'm in XX", write that city into the "Home City" field of user-rules.md so it applies automatically next time. Never modify hard rules unless the user explicitly says "change the rules".