Imported from tiangong-ai/agent-skills (openaq-data-fetch/SKILL.md). Install upstream with npx skills add tiangong-ai/agent-skills --skill openaq-data-fetch. Copyright stays with the author.

OpenAQ Data Fetch

Use the CLI-owned openaq.air-quality capability. This Skill supplies intent routing and result-use boundaries only; the CLI owns source discovery, input/output schemas, API-key injection, pagination, limits, validation, and receipts.

Before running

1. Read references/tiangong-data-requirement.json.
2. Use the caller- or workspace-resolved stable CLI. The requirement declares compatible capability and operation contract majors; it does not select a package build.
3. Run data describe with that same CLI. Continue only when the capability ID and required contract majors match, and copy the exact current capability/operation versions from that response into the run request.
4. Ensure OPENAQ_API_KEY is present in the CLI process environment, then run the default static doctor. Never place the key in argv, request JSON, a Skill-local file, logs, or output.

tiangong-ai data describe openaq.air-quality --json
tiangong-ai data doctor openaq.air-quality --json

Use the returned Discovery Metadata to confirm current ownership, coverage, granularity, source-specific quality and license limits, selection hints, provides, and doesNotProvide. Do not substitute facts remembered from an older Skill revision. A blocked static doctor means the required logical credential is unavailable; stop instead of bypassing the CLI.

Choose one operation

• Use search-locations when the relevant sensor ID is not yet known. Supply at least one explicit country, provider, parameter, license, monitor, mobile, center-radius, or bounding-box filter. Inspect provider, monitor status, parameter, license, and coverage metadata before choosing a sensor.
• Use fetch-sensor-measurements only after selecting one trusted sensor ID. Choose raw for upstream-reported observations or hourly/daily for OpenAQ's precomputed aggregates, and use an explicit RFC3339 window of no more than 366 days. Daily records use the sensor location's local-day boundaries; preserve the returned period instead of treating them as UTC calendar days.

Do not invent or pass an API path. Do not use this Skill to list or download OpenAQ S3 archive objects; bulk files require a separately governed content/download workflow. search-locations returns metadata attached to bounded location results; it is not a global country, provider, parameter, license, instrument, or sensor catalog.

Prepare a location request

Build a tiangong.data.run-request.v1 envelope and replace the version placeholders with the exact versions from the same data describe response. This example searches a bounded country/parameter combination; use the current input schema from data describe for other supported filters.

{
  "schemaVersion": "tiangong.data.run-request.v1",
  "capabilityId": "openaq.air-quality",
  "capabilityVersion": "<describe.manifest.capabilityVersion>",
  "operationId": "search-locations",
  "operationVersion": "<describe.manifest.operations[1].operationVersion>",
  "input": {
    "countryCode": "NL",
    "parameterIds": [2],
    "pageSize": 250,
    "sortOrder": "asc"
  }
}

Do not silently widen spatial or provider filters, turn a bounded search into a global scan, combine center with bounding box, geocode a place name, or infer that every returned sensor has continuous data for the desired time window.

Prepare a measurement request

Use the selected sensor's exact ID and an explicit granularity and time window. Do not merge sensors, convert units, or switch granularity inside this atomic call.

{
  "schemaVersion": "tiangong.data.run-request.v1",
  "capabilityId": "openaq.air-quality",
  "capabilityVersion": "<describe.manifest.capabilityVersion>",
  "operationId": "fetch-sensor-measurements",
  "operationVersion": "<describe.manifest.operations[0].operationVersion>",
  "input": {
    "sensorId": 1001,
    "granularity": "hourly",
    "startDateTime": "2026-03-01T00:00:00Z",
    "endDateTime": "2026-03-07T23:59:59Z",
    "pageSize": 1000
  }
}

Run

tiangong-ai data run openaq.air-quality search-locations \
  --input /absolute/path/to/location-request.json --json

tiangong-ai data run openaq.air-quality fetch-sensor-measurements \
  --input /absolute/path/to/measurement-request.json --json

Each command emits a tiangong.data.run-result.v1 envelope. Preserve its contract, warnings, errors, and receipt with data when handing the result to another workflow.

Result boundaries

• Preserve OpenAQ attribution and every original provider's license and attribution metadata. OpenAQ's terms do not replace source-specific terms.
• Treat location, owner, provider, instrument, monitor, sensor, and coverage fields as provider-dependent metadata, not proof of regulatory-network status or measurement quality.
• Keep raw, hourly, and daily granularities explicit. Do not mix raw upstream records with OpenAQ aggregates without an explicit analytical method, and retain flagInfo, aggregate summary, coverage interval, and period fields.
• Preserve provider nulls. A null value, period, coordinate, summary member, or coverage object is a valid unavailable field, not proof of zero, clean data, absence of pollution, or a failed request. Inspect flagInfo.hasFlags separately from nullability.
• Interpret daily aggregates in the selected location's timezone and preserve their returned period; do not relabel them as UTC calendar-day averages.
• Treat partial as incomplete page or record coverage. Report missing pages and provider errors with the usable records.
• Treat blocked as no usable business result. Surface credential, input, endpoint, rate-limit, or provider errors instead of bypassing limits, switching endpoints, or calling the old scripts.
• Report page/record-limit truncation. A truncated or empty result is not proof that no monitors or pollution observations exist.
• Do not calculate AQI, make health or regulatory claims, infer pollution sources, forecast conditions, interpolate space, correct calibration, convert units, or claim completeness from this result alone.
• Cross-sensor comparison, recurring monitoring, source fusion, statistical inference, bulk archive acquisition, and research evidence admission belong to the caller or Auto Research, not this atomic Skill.

Reference

• references/tiangong-data-requirement.json: stable capability requirement; it is not a package lock.