web-search

Pass

Audited by Gen Agent Trust Hub on Aug 1, 2026

Risk Level: SAFEDATA_EXFILTRATIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill transmits user-provided search queries to DuckDuckGo's public API to retrieve web, news, image, and video content.- [EXTERNAL_DOWNLOADS]: The skill depends on the duckduckgo-search Python package to interface with the DuckDuckGo search service.- [COMMAND_EXECUTION]: The script scripts/search.py allows search results to be saved to arbitrary local file paths via the --output argument, including the automatic creation of parent directories.- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface:
  • Ingestion points: Untrusted web data is fetched from DuckDuckGo in scripts/search.py via ddgs.text(), ddgs.news(), ddgs.images(), and ddgs.videos().
  • Boundary markers: No delimiters or warnings are used to separate untrusted search content from system or user instructions.
  • Capability inventory: The skill can write search results to the file system using Path(args.output).write_text().
  • Sanitization: No content filtering or sanitization is performed on the results retrieved from the web before they are displayed or saved.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 1, 2026, 09:26 AM
Security Audit — agent-trust-hub — web-search