SCBE Connector Health Check

Use this skill when connector reliability matters before running social research, outreach, or automation pipelines.

Operating Contract

1. Verify capability availability before building assumptions.
2. Record service state as ok, requires_auth, degraded, or down.
3. Prefer minimal required scope checks first, then deeper verification.
4. Never claim capability if tool access is not confirmed.
5. Always return explicit next action for each degraded connector.

Scope

Use this for:

• checking MCP server/tool availability
• verifying API auth and endpoint reachability
• validating connector prerequisites for marketing/outreach workflows
• periodic health snapshots for SCBE runbooks

Quick Health Workflow

1. Inventory connectors

   • list known local connector groups and expected services.

2. Availability probes

   • check tool discoverability and response latency.

3. Auth & permission check

   • confirm required env vars/tokens are configured.
   • verify command-level access for critical calls.

4. Functionality spot-check

   • run one non-destructive call per connector (read/list/search only).

5. Risk annotation

   • tag degraded services with:
     • impact level (none/low/medium/high)
     • fallback behavior
     • owner and triage owner if available

6. Publish health record

   • provide compact checklist and ordered repair steps.

Output Contract

• inventory: expected services + owner intent.
• status: per-connector status map.
• checks: quick probe results and timestamps.
• failures: auth or dependency blockers with exact error text.
• next_steps: prioritized remediation commands or decisions.
• ready_for_live_use: true only if primary services are green.

Connector Priority Matrix (example)

• primary: GitHub, Notion, Obsidian, HF, browser service.
• secondary: Linear, Discord/Slack, Telegram, marketplace tools.
• tertiary: optional APIs used only for enrichment.

Recovery Patterns

• If requires_auth: document missing scopes, rotate creds, rerun spot-check.
• If degraded: reduce non-critical calls, queue retries, add backoff and cache.
• If down: disable workflow branch and use fallback runbook with explicit user visibility.

References

• references/connector-health-checklist.md