| name | weight-gain-strategy |
| version | 1.0.0 |
| description | Deep intervention for sustained upward weight trends. Trigger: consecutive increases flagged by post-weigh-in checks, or explicit questions ('体重怎么涨了', 'why am I gaining weight'). Graduated support: reassurance → cause analysis → temporary adjustments. Recording weigh-ins and routine trend judgment → weight-tracking. NOT when emotional distress needs priority → emotional-support; NOT when weight-focus should be avoided (history_of_ed / avoid_weight_focus flags). |
| metadata | {"openclaw":{"emoji":"mag","homepage":"https://github.com/NanoRhino/weight-loss-skill"}} |
Weight Gain Strategy
Detect upward weight trends and respond with graduated support — from
reassurance on the first increase, to guided cause discovery, to full
diagnosis with adjustment strategies — matching the response depth to how
persistent the trend is.
Routing Gate
Entry paths:
- Auto (post-weigh-in):
weight-tracking 记完体重后自主判断是否需要干预 → 需要时读 references/cause-check-flow.md 进入诊断流程。
- Manual: User asks about weight gain ("why am I gaining weight", "体重怎么涨了") → check Skip conditions first → run
analyze directly → Interactive Flow Step 1.
Skip — do NOT enter this skill if:
- No
PLAN.md exists (no plan to deviate from)
USER.md > Health Flags contains avoid_weight_focus or history_of_ed- User shows emotional distress about weight → defer to
emotional-support (P1 priority)
Principles
- Normalize first. Lead with reassurance, then dig into data.
- Data + habits before opinions. Every diagnosis must cite actual numbers or observable behavioral patterns. Never speculate without evidence.
- Escalate gradually. Response depth follows the streak. Never skip levels or jump to strategy on a first increase.
- Collaborate, don't force. The user can opt in or out at every step. Playful challenges are fine; pushing past a "no" is not.
- Keep it light. Witty friend, not stern doctor. Data rigorous, delivery fun.
Diagnosis Dimensions
The analyze command outputs raw statistics only — no detected: true/false judgments. The AI interprets these numbers in context (user history, lifestyle, chat context) to determine causes.
Output fields
| Field | What it contains | AI uses it for |
|---|
calorie_stats | avg/min/max/std_dev, days over target, days under 60%, daily breakdown | Surplus, volatility, binge/restrict patterns |
protein_stats | avg daily g, recommended g (weight×1.2), days below 70% | Protein deficit detection |
exercise_stats | This week vs last week sessions & minutes | Exercise decline |
logging_stats | Coverage %, single-meal days, unlogged days | Data reliability |
weight_pattern | Largest daily jump + dates | Sudden spike (water retention) |
food_list | Raw food names (dedupe, up to 50) | Food quality, variety, processed patterns |
data_confidence | sufficient flag, issues list | Whether to analyze or ask for more data first |
active_strategy | Current strategy type/dates if active | Whether to suppress new interventions |
suggested_actions | Concrete script-driven actions (not AI judgment) | Strict mode, set calorie target, suppress strategy |
⚠️ suggested_actions are deterministic rules, not AI opinions:
strict_mode: coverage < 50% or >50% single-meal days → enter strict mode (see references/strict-mode.md). Do NOT create new meal reminder crons — they already exist. Strict mode makes existing reminders more insistent.set_calorie_target: no calorie target set → cannot do surplus analysissuppress_new_strategy: active strategy hasn't expired → don't start a new cause-check
🎯 AI creates targeted habits based on analysis — NOT generic meal reminders:
After analyzing the raw data, the AI identifies the specific problem and creates a habit that addresses it. Examples:
- Protein low → habit: "每餐加一份蛋白质（鸡蛋/鸡胸/豆腐）"
- Calorie volatility (binge/restrict) → habit: "每天吃到{目标}附近，不跳餐"
- Late-night eating pattern → habit: "8点前吃完晚饭"
- Weekend overeating → habit: "周末拍照打卡，不多不少"
- Food quality issues → habit: specific swap based on actual foods (e.g. "方便面换成挂面煮蛋")
- Snacking excess → habit: specific swap (e.g. "下午零食换成酸奶/坚果")
NEVER create habits for: meal logging reminders, weight check-ins, or anything that already has a cron job.
⚠️ AI-driven analysis: The script provides numbers; the AI decides what they mean. A std_dev of 967 kcal might be binge/restrict — or a user transitioning diets. The AI considers context.
Analysis Script
Script path: python3 {baseDir}/scripts/analyze-weight-trend.py
Commands: analyze, save-strategy, check-strategy.
See references/script-api.md for full usage, parameters, and return schemas.
Safety Rules
- Calorie floor: Never suggest intake below max(BMR, 1000 kcal/day).
- Exercise safety: For sedentary users or those with health conditions, start with walking only.
- No shame, no blame. Frame adjustments as experiments, not corrections.
References
| File | Contents |
|---|
references/cause-check-flow.md | Full cause-check flow (Steps A→D), habit creation, cron rules |
references/script-api.md | Script commands, parameters, return schemas |
references/strict-mode.md | Strict mode: trigger, behavior rules, duration, failure escalation |
references/data-schemas.md | Data sources, strategy JSON schema, skill integration, edge cases |
