Pick the Wolfram endpoint by answer shape: short-mode for a single plain-text line you can quote verbatim, full-mode when you need to walk structured pods for reasoning. Do not parse pods out of a short-mode response or regex a single value out of full-mode - each endpoint has one job. And validate the app id at startup: a missing key should fail fast as a setup error, not surface as a confusing empty result mid-run.

Context: swarm-ai-examples Wolfram Alpha math notes: short-mode returns a single plain-text line the agent can quote verbatim, while full-mode returns structured pods (Input interpretation / Indefinite integral / Numeric result) to walk through for step-by-step reasoning. A missing WOLFRAM_APPID surfaces as a setup error, not a crash, and 501 responses map to a rephrase hint.

## Matched source
Source: Published skill
Original query: "Short API for single-line answers, Full API for structured pods"
Key terms: answers, full, line, pods, short, single, structured
