airshell_skill

Overview

This skill provides an agent playbook for the AirShell air quality sensor, giving your agent domain knowledge about CO₂, PM2.5, temperature, and humidity. It guides a setup flow to connect a Sensirion SEN63C + Raspberry Pi sensor to your agent via webhook and defines decision logic so the sensor wakes the agent only when air quality needs attention. The agent interprets readings against occupant-sensitive thresholds and gives concise, context-aware advice.

How this skill works

On first run the skill runs a setup interview to capture room, occupants, location, priorities, and notification preferences, then writes a deployment file with agreed thresholds. After setup the skill reads the deployment context, fetches sensor status and recent trends, checks outdoor weather before giving ventilation advice, and decides whether an alarm merits notifying the user. It can push configuration to the device, manage optional purifier control scripts, and tailor urgency based on occupant sensitivity.

When to use it

• When onboarding a Sensirion SEN63C + Raspberry Pi AirShell device to your agent
• When an AirShell alarm (RAISED / REPEAT / CLEARED) arrives and needs interpretation
• When a user asks for current air quality or recent trends from a room sensor
• When adjusting alarm thresholds based on occupant sensitivity or comfort priorities
• When automating purifier speed changes on PM2.5 events (if purifier configured)

Best practices

• Complete the setup interview one question at a time and store deployment context before processing alarms
• Always read references/deployment.md for room, occupants, location, and agreed thresholds before interpreting readings
• Check outdoor weather/PM2.5 before recommending opening windows to avoid making indoor air worse
• Keep notifications short: state the measured problem, relevance to occupants, and a single actionable step
• Push configs including gateway webhook and token to the device; handle missing webhook/token by instructing how to find gateway config

Example use cases

• Nursery CO₂ rises: agent checks deployment (infant), fetches device status, and sends a short urgent ventilation prompt
• Home office PM2.5 spike: agent checks outdoor PM2.5 and suggests opening windows only if outdoor air is cleaner
• User requests a 2-hour trend: agent queries recent readings and summarizes whether metrics are stable, improving, or worsening
• Adjusting thresholds: agent discusses standards, confirms changes, updates deployment file, and POSTs new config to the device
• PM2.5 alarm triggers purifier control script: agent runs the configured command silently and logs failures