bing-search
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local bash script (
scripts/search_bing.sh) to perform its primary function. This script utilizescurlfor network communication andpython3for structured data validation and processing. - [INDIRECT_PROMPT_INJECTION]: As a search tool, the skill naturally ingests untrusted data from an external source (Bing search results) which could contain adversarial content.
- Ingestion points: Search results are retrieved from the Frevana API at
https://ai-factory.frevana.com/service/bing-search(inscripts/search_bing.sh). - Boundary markers: The script validates that the API response is valid JSON before outputting or saving it, though it does not provide natural language delimiters.
- Capability inventory: The skill has the ability to write search results to the filesystem via the
--outputparameter. - Sanitization: The content within the JSON payload (search snippets, titles) is not sanitized for potential prompt injection instructions.
Audit Metadata