The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses curl in both documentation examples and a dedicated shell script (scripts/searxng.sh) to interact with SearXNG APIs. The shell script uses a basic sed command to replace spaces with plus signs in the query, which is a fragile form of URL encoding and does not account for other special characters that might affect the query string structure.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external search engines, which represents a surface for indirect prompt injection.
Ingestion points: Search result titles, snippets, and URLs are fetched from the SearXNG API in SKILL.md and scripts/searxng.sh.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the ingested search snippets as untrusted data.
Capability inventory: The skill has access to terminal execution for curl and python3, and the workflow suggests further processing of results via web extraction tools.
Sanitization: JSON responses are parsed using standard libraries, but the text content within those results is not sanitized for potential malicious instructions before being passed to the agent.
[DYNAMIC_EXECUTION]: The skill's documentation provides examples of using python3 -c to execute small inline scripts for parsing JSON output from curl. This is a form of dynamic execution used for local data processing.

searxng-search