Dependency outage triage

Use OutageDeck's public MCP tools to establish vendor evidence before changing code. Treat status data as one diagnostic signal, not proof of causation.

Workflow

1. Identify the external providers implicated by the request, logs, configuration, or stack description. Do not expose secrets while inspecting evidence.
2. Resolve an ambiguous product or company name with search_providers. Do not guess a provider slug after a failed lookup.
3. Check current status:
   • Use get_provider_status for one provider.
   • Use one check_my_stack call for 2-12 providers. Split larger stacks into batches of at most 12.
   • Use list_active_incidents only for an ecosystem-wide question such as "what is down right now?"
4. When a result includes an incident, call get_incident_details with its exact slug. Use search followed by fetch when the user needs a citable document or the incident must be found from descriptive text.
5. Use get_uptime for one provider's 7-90 day history. Use get_outage_report for a cross-vendor historical comparison. Do not use historical uptime to claim that a current failure is vendor-caused.
6. Compare the vendor timeline, affected service, and severity with the user's error and timestamps.
7. Return the evidence and the narrowest defensible verdict before proposing code changes.

Output contract