toddler-meltdown-decoder

Use during or after a toddler meltdown (age ~1-4) — triages the trigger class from observed cues (hungry, tired, overstimulated, frustrated, sensory, connection-seeking, pain) with a ranked checklist, hands you the right de-escalation script per class and the words NOT to use mid-meltdown, flags pediatrician-level red flags (self-injury, breath-holding, total inconsolability), and keeps a pattern log that finds your child's time-of-day hotspots and recurring trigger chains so meltdowns get prevented, not just survived.

A melting-down toddler is not a discipline problem — it's a child with an arrested prefrontal cortex and one unmet need, and a parent doing detective work at 100 dB. The failure mode isn't lack of love; it's applying the WRONG protocol: reasoning with a screaming 2-year-old (language centers offline), punishing the hunger tantrum (feeds the spiral), or missing the pattern that every meltdown is 17:30 ± 20 minutes.

This skill is the triage card + pattern log:

1. Triage — score 7 trigger classes from the cues you can actually observe (no mind-reading required): hungry, tired, overstimulated, frustrated/skill-mismatch, sensory, seeking connection, pain/teething. Ranked output names the top class and its confidence.
2. De-escalate — per-class script: body position, the 3-5 words that work, what to offer, and the class-specific NEVER list ("don't offer choices to an overstimulated child" vs "always offer two tiny choices to a frustrated one").
3. Flag — red flags that are pediatrician conversations, not parenting failures.
4. Log & prevent — every episode logged with time, cues, class; the stats view finds hour-of-day hotspots and the most common trigger chain, which is where prevention lives (snack at 17:00 beats perfect de-escalation at 17:40).

When to Use

• Mid-meltdown, phone in hand: run triage --cues ..., get the script in 10 seconds
• Post-meltdown debrief (the honest moment): log it
• Weekly: stats to see the pattern — hotspots, recurring classes, whether the 17:00 snack intervention worked
• Second child behaving completely differently: re-log from zero, classes stay the same but cue weights are per-child
• Caregiver handoff: print the child's top-3 classes and their scripts for grandparents/daycare

Don't use for: babies under ~12 months (cry decoding is a different problem — see feeding/sleep skills), school-age behavior plans, or suspected injury/illness as the PRIMARY suspect (pain class routes you to checking, then to care).

The Seven Classes (cue signatures)

Class	Cue signature	Onset shape
HUNGRY	pre-meal timing, "hangry" aggression, cries then bites/accepts food fast	fast ramp
TIRED	eye rub, yawns, increased clumsiness, melts at minor frustration that morning-self handled	slow burn, worse fighting it
OVERSTIMULATED	covering ears, avoiding eye contact, wants to be left alone, busy environment just before	escape-seeking
FRUSTRATED	mid-task collapse (block tower, clothes, toy), "me do it", throws the object AT the problem	task-linked
SENSORY	tags/clothing/seams, sock seams, sounds, lights, transitions between textures/temperatures	specific stimulus
CONNECTION	parent on phone/another child, follows then hits/bites, calms FAST with attention	bid-then-boom
PAIN/TEETHING	chewing on things, ear-grabbing, drooling, wakes at night, nothing else works	background hum + spikes

Classes stack (tired+hungry is the classic 17:30 double). The triage output shows the top TWO with weights.

De-escalation Doctrine (all classes)

• Get low — below eye level, sideways posture, not looming
• 3-5 words max — "You're safe. I'm here." Sentences don't land; presence does
• Name the feeling once — "So mad." Labeling helps AFTER the peak, not during
• Fewer stimuli — dim, quiet, fewer people, fewer words
• No teaching mid-meltdown — the lesson ("we don't hit") comes at reconnection, 10+ min later, in one sentence
• Time-IN not time-out under ~4 — isolation escalates; co-regulation is the mechanism that builds self-regulation

Pitfalls

1. Reasoning mid-meltdown. Language processing is offline; every extra word is more stimulus. Presence > sentences.
2. The bribery trap. Treat/toy to stop the tantrum teaches the tantrum, fast. Meet the NEED (hunger, quiet, connection), never reward the volume.
3. Punishing the HUNGRY class. A food-tantrum punished is a child who learns needs are dangerous. Feed first, always.
4. Asking "what's wrong?" Open questions demand language the child doesn't have under stress. Offer binary choices instead — except for the overstimulated class, where even choices are load.
5. Logging only the meltdowns. Log the near-misses too (--averted) — they identify which interventions work.
6. Ignoring your own state. Parent dysregulation is the #1 escalator. The script prints a one-line parent check: you can't co-regulate while dysregulated — tag out if possible.
7. Chronic-pattern blindness. 3+ meltdounds/day every day for weeks isn't a discipline issue — see red flags.

Red Flags → Talk to a Pediatrician

• Self-injury (head banging beyond brief frustration, biting self hard)
• Breath-holding spells that cause fainting
• Total inconsolability — NOTHING works, every time
• No meltdowns ever responding to connection (contrast: typical toddlers calm with a trusted adult within minutes)
• Language regression or absence of pointing/joint attention alongside the behavior
• Nightly 45+ minute episodes with aggression

None of these are diagnoses — they're the "worth a conversation" list the triage output surfaces when logged cues match.

Verification Checklist

• Triage run from OBSERVED cues (not guessed feelings)
• Top class script read fully, including its NEVER list
• Episode logged with time + cues + class + duration
• Weekly stats reviewed; hotspots addressed upstream (snack/schedule/environment), not mid-crisis
• Red-flag cues checked against the list, not rationalized
• Caregiver brief (top-3 classes + scripts) shared

Quick Recipe

text

# mid-meltdown: 10-second triage
python3 scripts/meltdown_triage.py triage \
    --cues "rubbing-eyes,aggression,17:40,clumsy" --age 2

# debrief: log what actually happened
python3 scripts/meltdown_triage.py log --age 2 \
    --cues "ear-grab,chewing,night-waking" --class pain \
    --duration 20 --notes "molar #3"

# weekly pattern review
python3 scripts/meltdown_triage.py stats