yuga-hashimoto/open-dash/breathing-exercise

Breathing Exercise

Trigger on "breathing exercise", "box breathing", "4 7 8 breathing", "深呼吸", "呼吸法". Distinct from meditation (longer, open-ended, silent) and stretch-break (physical movement): this is a tight, voice-guided ~1-minute breathwork drill using the 4-7-8 pattern.

Each cycle: inhale 4 seconds → hold 7 seconds → exhale 8 seconds. Run 4 cycles (~76 seconds of breathing plus short intro/outro).

Default flow

1. Volume check: if user is in quiet-hours (see quiet-hours

skill), keep set_volume at 20 or below. Otherwise set to 30 — audible but unobtrusive.

1. Intro: speak one sentence: "Four cycles of four-seven-eight

breathing. Follow my cues." Then pause 2 s before starting.

1. Cycle loop (repeat 4 times):

• Say "Breathe in" at the start of inhale.
• set_timer { seconds: 4, label: "breath-inhale" } — when it

fires, say "Hold".

• set_timer { seconds: 7, label: "breath-hold" } — when it

fires, say "Breathe out slowly".

• set_timer { seconds: 8, label: "breath-exhale" } — when it

fires, pause ~500 ms before next cycle.

• Between cycles, add a one-beat "Again." cue so the user knows the

next inhale is starting.

1. Outro: after the 4th exhale timer fires, say once: "Done. Nice

work." Do not congratulate beyond that — keep it calm.

Variants

• Box breathing (4-4-4-4): if the user says "box breathing"

explicitly, use 4-second inhale, 4-second hold, 4-second exhale, 4-second hold. Run 4 cycles (~64 s total). Add a "Hold empty" cue for the post-exhale hold.

• Longer session: if user says "keep going" / "もう一回" during

outro, run another 4 cycles without re-intro.

Interrupt handling

If the user says "stop" / "やめる" / "もういい" mid-session:

1. cancel_all_timers scoped to the breathing-exercise labels.
2. Speak once: "Okay." — no judgement, no recap.

Quiet-hours respect

If quiet-hours is active:

• Never rise above volume 20.
• Skip the intro; just whisper "In", "Hold", "Out" instead of full

sentences.

• No outro — end in silence.

Multi-room

Do NOT broadcast_tts or broadcast_announcement. Breathing is personal and the timing cues must match the caller's breath. If a household wants a synchronised session they'll explicitly ask, and even then — decline and suggest they each start their own.

Style

• Monk-voice adjacent, like meditation: quiet, short, steady.
• Never cheerful. Never count aloud ("one… two…") — the timers do

that work. Cue words only.

• JA: polite-casual, breath-length ("吸って", "止めて", "吐いて").

Tools used

• set_timer (inhale / hold / exhale, short labels)
• cancel_all_timers (on interrupt or at end)
• set_volume

Tools explicitly avoided

• broadcast_tts / broadcast_timer / broadcast_announcement —

breath cues are individual and must not leak to other rooms.

• execute_command for lights — the user's eyes are likely closed;

don't fiddle with the room state for a 1-minute drill.

• morning_briefing / evening_briefing — noise.