The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the official SerpApi CLI via Homebrew and references official SDKs for Python, JavaScript, Ruby, Go, PHP, Java, and .NET. These resources are hosted on official package registries and the vendor's GitHub organization.
[COMMAND_EXECUTION]: Contains documentation for utilizing the serpapi CLI tool and curl to query search results and manage account status. The commands are standard for the service's functionality and include proper authentication procedures.
[INDIRECT_PROMPT_INJECTION]: As a search tool, the skill ingests and processes untrusted data from the open web (search engine results). This creates a vulnerability surface for indirect prompt injection, where an attacker could embed malicious instructions in a web page to influence the agent's behavior during result processing.
Ingestion points: Untrusted data enters the agent context through the search results returned by various engines (e.g., google_light, google_news) as documented in SKILL.md and rules/ENGINES.md.
Boundary markers: The skill does not explicitly define prompt boundaries, though the use of structured JSON results and the mode="compact" parameter in the MCP tool provides a level of data-instruction separation.
Capability inventory: The skill facilitates network operations to serpapi.com and provides access to local execution of the serpapi binary as defined in serpapi.yaml.
Sanitization: The skill encourages the use of jq filters and the --fields parameter to extract specific machine-readable keys (e.g., .title, .link), which limits the amount of untrusted text processed by the agent.

serpapi-web-search