The Agent Skills Directory

[SAFE]: The skill provides defensive architectural patterns for integrating the SerpAPI SDK, focusing on creating a secure gateway that prevents vendor-specific details from leaking into business logic.
[SAFE]: It explicitly enforces security best practices by instructing users to inject credentials via environment variables (SERPAPI_KEY) and prohibits their inclusion in logs, telemetry, or error objects.
[SAFE]: The instructions recommend implementing redaction and error normalization to prevent sensitive information disclosure during failures.
[INDIRECT_PROMPT_INJECTION]: The skill describes a pattern for processing external search results, which represents a potential surface for indirect prompt injection.
Ingestion points: External search data is retrieved via the SerpAPI client and ingested into the application context.
Boundary markers: The skill mandates the use of a "normalized response type" and "typed, testable boundaries" to delimit vendor data.
Capability inventory: The skill uses WebFetch for interface verification and Write/Edit for adapter implementation.
Sanitization: The patterns specifically include "redaction," "sanitized fixtures," and "contract tests" to validate and clean external inputs.

serpapi-sdk-patterns