search
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration were detected. The skill uses environment variables to manage credentials, which is a standard security practice.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external web sources (Google search results) which could potentially contain adversarial instructions intended to influence the agent's behavior.
- Ingestion points: Search result titles, links, and descriptions fetched from the web in
scripts/search.sh. - Boundary markers: The output is structured as JSON, but individual strings (descriptions) do not contain explicit "ignore instructions" delimiters.
- Capability inventory: The skill uses
curlfor network requests andjqfor JSON processing inscripts/search.sh. - Sanitization: The user-provided search query is URI-encoded using
jqto prevent URL manipulation.
Audit Metadata