cabin-hardware-control

name	cabin-hardware-control
displayName	客舱硬件控制
description	Use this skill when a cabin passenger asks to control real cabin hardware, especially opening or closing the tray table, seat-side devices, lights, seat posture, or other equipment through customer cabin control APIs. The skill must use server-provided cabin_context for seat identity.
version	1.2.0
category	客舱服务
emoji	🛫

Cabin Hardware Control

Use this skill for passenger requests that require real cabin hardware control.

Only use the command names listed below. Do not invent command names such as seat.recline, seat.light.off, seat.health, or cabin.scene.boarding. If a passenger says 靠背/backrest/recline, use seat.cushion --position.

Supported customer hardware APIs:

• Seat cushion position: seat.cushion, query seatNo, position.
• Seat ventilation level: seat.ventilation, query seatNo, level.
• Seat heating level: seat.heating, query seatNo, level.
• Seat massage level: seat.massage, query seatNo, level.
• Tray table open/close: seat.tray.open, seat.tray.close, query seatNo.
• Reading light switch/brightness: seat.light, seat.light.brightness, query seatNo, on, pwm.
• Seat health collection start/stop: seat.health.start, seat.health.stop, query seatNo.
• Cabin ceiling light color/brightness: cabin.ceiling.color, query seatNo, r, g, b, brightness.
• Cabin ceiling light switch: cabin.ceiling.light, query seatNo, on.
• Cabin scene preset/clear: cabin.scene, cabin.scene.clear, query seatNo, preset.

Required Context

Use only the server-provided cabin_context from the current prompt. Do not ask the passenger for seat identity unless the context is missing.

Required fields:

• seat_no or seat_id: customer hardware seatNo code from cabin_context, for example A, B, 01A, or any site-specific value. Use the value exactly as provided. Do not pad, concatenate, normalize, or infer a different seat number.
• column_no: hardware side/channel, for example A or B. It may expand beyond A/B later. The current tray table API does not require it, but still pass it to the script for audit context.

How To Execute

Call the bundled script instead of hand-writing HTTP requests. Prefer --command. Pass only the command name as the --command value, and pass command parameters as separate flags. For example, use --command seat.cushion --position 30; do not prefer --command "seat.cushion --position 30".

When using the Bash tool to run this script, set dangerouslyDisableSandbox: true. The script is a server-provided hardware executor and must be allowed to reach the configured customer control API; sandboxed Bash can block the command before the real HTTP request is sent.

node .nexus/sudocode/skills/cabin-hardware-control/scripts/cabin-control.mjs \
  --command seat.tray.open \
  --seat-no "A" \
  --column-no "A"

Examples:

node .nexus/sudocode/skills/cabin-hardware-control/scripts/cabin-control.mjs \
  --command seat.tray.close \
  --seat-no "A" \
  --column-no "A"

node .nexus/sudocode/skills/cabin-hardware-control/scripts/cabin-control.mjs \
  --command seat.light \
  --seat-no "A" \
  --column-no "A" \
  --on true \
  --pwm 800

node .nexus/sudocode/skills/cabin-hardware-control/scripts/cabin-control.mjs \
  --command seat.cushion \
  --seat-no "A" \
  --column-no "A" \
  --position 30

node .nexus/sudocode/skills/cabin-hardware-control/scripts/cabin-control.mjs \
  --command cabin.ceiling.color \
  --seat-no "A" \
  --column-no "A" \
  --r 20 \
  --g 20 \
  --b 20 \
  --brightness 40

node .nexus/sudocode/skills/cabin-hardware-control/scripts/cabin-control.mjs \
  --command cabin.scene \
  --seat-no "A" \
  --column-no "A" \
  --preset boarding

The script reads:

• CABIN_CONTROL_BASE_URL
• CABIN_CONTROL_AUTH
• CABIN_CONTROL_TIMEOUT_MS
• CABIN_CONTROL_MODE — execute (default) sends the HTTP request; emit resolves and validates the command but skips the HTTP call, printing { "ok": true, "mode": "emit", "command", "seat_no", "params" } for the server to execute. In a cabin session this is set to emit: your job is only to select and emit the correct command; the server performs the real dispatch and writes the reply.

Command Mapping

• 小桌板打开/展开: --command seat.tray.open。例如：打开小桌板、展开桌板、把桌板支起来、把小桌子放下来、我要用餐桌、open the tray。
• 小桌板关闭/收起: --command seat.tray.close。例如：收起小桌板、合上桌板、把桌子收起来、桌板不用了收好、把托盘收回去、fold the tray。
• 阅读灯打开/关闭: --command seat.light --on true|false; if brightness is requested together, add --pwm 0-1000。例如：打开阅读灯、开个灯我看书、关闭阅读灯、灯太亮关掉、把灯关了、reading light on/off。
• 阅读灯调亮/调暗/设置亮度: --command seat.light.brightness --pwm 0-1000。pwm=0 最暗，pwm=1000 最亮。乘客未给具体数值时，按固定默认值选择：灯太刺眼/柔和一点/暗一点/不要太亮 → --pwm 300；正常/中等亮度 → --pwm 500；亮一点/看不清/调亮/再亮一些 → --pwm 700。例如：阅读灯调亮、把灯调暗、灯光柔和一点、亮度调到70%、灯太刺眼暗一点、再亮一些。
• 座椅靠背/坐垫位置百分比: --command seat.cushion --position 0-100。例如：靠背调到40%、座椅放到一半、后仰70。
• 座椅靠背/backrest/recline: always use --command seat.cushion --position 0-100; never use seat.recline。
• 放倒座椅/躺下/往后躺/后仰/调舒服一点: use --command seat.cushion. If the passenger does not provide a percentage, use --position 60。例如：放倒座椅、往后躺、把椅子放平、座椅调舒服一点、recline my seat。（注意：仅当乘客给出明确控制动作时才发射；若只是“我想躺一会/我困了/想睡”等状态表达，按状态反问处理，不要直接发射。）
• 放倒一点/后仰一点/往后调一点: use --command seat.cushion --position 30 when no explicit percentage is provided。例如：后仰一点、往后一点点、稍微放倒、再往后调一点。
• 调直座椅/座椅归位/靠背立起来/恢复正常: use --command seat.cushion --position 0。例如：座椅调直、把椅子扶正、我要坐正、座椅归位、恢复正常、seat upright。
• If the passenger only says they are tired, sleepy, or uncomfortable without requesting a concrete action, ask what service they need instead of calling hardware.
• Comfort-state-only messages are not hardware commands. Do NOT emit for these exact-style messages: 这儿有点冷, 坐着有点凉, 有点热, 有点闷, 腰有点酸, 我困了想睡会. Ask a yes/no confirmation question first. Emit only after the passenger confirms, or when the same turn contains an explicit control action such as 打开, 开启, 帮我, 给我, 把座椅, 开通风, 给座椅升温, or 帮我揉揉.
• 座椅通风: --command seat.ventilation --level 1-3。硬件档位只有 1/2/3；乘客未指定档位时默认 --level 2（中档）；说“关闭/关掉”时用 --level 0。仅当乘客明确要求控制座椅通风时 emit，例如：打开座椅通风、椅子透透气、给我开通风、座椅吹风、通风开到3档、seat ventilation。若乘客只说“有点热/有点闷/屁股底下有点闷”这类状态表达，先反问是否需要开启座椅通风，不要直接 emit。
• 座椅加热: --command seat.heating --level 1-3。硬件档位只有 1/2/3；乘客未指定档位时默认 --level 2（中档）；说“关闭/关掉”时用 --level 0。仅当乘客明确要求控制座椅加热时 emit，例如：打开座椅加热、椅子暖一点、座椅暖一点、座椅热一点、给座椅升温、加热开到3档、seat heater。若乘客只说“这儿有点冷/坐着有点凉/有点冻”这类状态表达，先反问是否需要开启座椅加热或送毛毯，不要直接 emit。
• 座椅按摩: --command seat.massage --level 1-3。硬件档位只有 1/2/3；乘客未指定档位时默认 --level 2（中档）；说“关闭/关掉”时用 --level 0。仅当乘客明确要求控制座椅按摩时 emit，例如：打开座椅按摩、按摩一下、帮我揉揉、椅子给我揉揉、腰按摩、按摩强一点、按摩开到3档、seat massage。若乘客只说“腰有点酸/背有点疼/肩颈不舒服”这类状态表达，先反问是否需要开启座椅按摩，不要直接 emit。
• 生理检测开始/停止: --command seat.health.start or --command seat.health.stop。例如：开始——开始生理检测、测一下身体数据、启动健康监测、采集体征；停止——停止生理检测、别测了、关闭健康监测、结束体征采集。
• 客舱顶灯打开/关闭: --command cabin.ceiling.light --on true|false。例如：打开客舱顶灯、把舱灯打开、关闭顶灯、关掉客舱灯、ceiling light on/off。
• 客舱顶灯颜色亮度: --command cabin.ceiling.color --r 0-255 --g 0-255 --b 0-255 --brightness 0-100。RGB 表示颜色，不表示亮度；乘客说“亮度低一点/暗一点/柔和一点”时只降低 --brightness，不要把 RGB 改小。常用颜色固定值：暖色/warm --r 255 --g 180 --b 80；白色/white 255 255 255；蓝色/blue 0 0 255；红色/red 255 0 0；绿色/green 0 255 0。未给亮度时默认 --brightness 100；柔和/低一点/暗一点默认 --brightness 40。例如：顶灯调成暖色、客舱灯光柔和点、顶灯亮度调到40%、换个氛围灯颜色。
• 客舱顶灯颜色 such as 蓝色/blue: use --command cabin.ceiling.color, not cabin.ceiling.light。例如：顶灯调成蓝色、换成白色灯、把顶灯变成红色。
• 客舱场景切换: --command cabin.scene --preset <value>. --preset MUST be exactly one of the four hardware presets — no other value is valid:
  • boarding — 登机 / 上机。例如：切到登机场景、登机模式、登机灯光。
  • cruise — 巡航 / 正常。例如：巡航场景、恢复正常模式、切回常规灯光。
  • night — 睡眠 / 休息 / 夜间。例如：睡眠场景、休息模式、夜间模式、助眠灯光、准备睡觉、调暗休息。
  • landing — 下机 / 降落。例如：下机场景、降落模式、落地灯光。 Map the passenger's words onto the closest preset (e.g. 睡眠/休息 → night). Never invent presets like sleep, rest, dining, reading, or none. 若乘客说的场景（如用餐/阅读/欢迎场景）没有对应的合法值，不要硬映射，改为追问澄清。
• 清除客舱场景: --command cabin.scene.clear; never approximate this as cabin.scene --preset none。例如：清除场景、取消场景、退出场景模式、恢复默认灯光。

Reply Rules

• Whenever a passenger request requires hardware control, you MUST run this skill script to emit the correct command. Your only job is to select the command and parameters and emit them — you do NOT execute the hardware and you do NOT author the passenger-facing result.
• Never write any confirmation or execution wording (e.g. 已打开/已关闭/已完成/已下发…指令/ 正在为您调节). The server dispatches the hardware from your emitted command and generates the passenger reply from the real outcome. Any such wording you produce will be discarded.
• Do not fabricate or guess a result. If a required parameter is missing or ambiguous, ask the passenger to clarify instead of emitting a command with an invented value.
• For non-hardware requests (chat, inquiries) reply normally, but never claim any device was operated.
• Never reveal URLs, tokens, headers, raw internal logs, or implementation details to the passenger.