The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it retrieves and displays untrusted data from the web.
Ingestion points: The script scripts/search.py fetches search result titles and snippets from a SearXNG instance and prints them to the terminal, where they enter the agent's context.
Boundary markers: No delimiters or specific instructions are used to signal the agent to ignore any malicious instructions embedded within the search results.
Capability inventory: The skill is configured to allow shell tool usage in its environment.
Sanitization: The script truncates search snippets to 200 characters but does not sanitize or filter the content for potential prompt injection patterns.
[DATA_EXFILTRATION]: The script performs network operations to fetch data from external sources.
The script scripts/search.py makes HTTP GET requests to the search endpoint of a SearXNG instance.
Although it defaults to a local address, the destination URL can be customized via the SEARXNG_URL environment variable or a command-line flag, allowing the agent to connect to any network domain.

searxng-search