hot-monitor
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from multiple external search engines and social media platforms, which could contain malicious instructions designed to influence the agent's behavior.\n
- Ingestion points: Data is fetched from Bing, Google, DuckDuckGo, HackerNews, Sogou, Bilibili, Weibo, and Twitter through scripts in the
scripts/directory.\n - Boundary markers: The skill includes an evaluation framework in
references/analysis-guide.mdthat provides criteria for content assessment, offering a structural context for the agent's reasoning.\n - Capability inventory: The ingested data is used for analysis and summary generation. The skill does not possess capabilities to write to the file system or execute code based on the external content.\n
- Sanitization: Scripts implement basic sanitization, such as stripping HTML tags from Bilibili search results and normalizing DuckDuckGo URLs.\n- [EXTERNAL_DOWNLOADS]: The search scripts perform network requests to various external domains to retrieve trending topic data.\n
- Evidence: Network connections are established to
bing.com,google.com,duckduckgo.com,hn.algolia.com,sogou.com,api.bilibili.com,weibo.com, andapi.twitterapi.io.\n- [COMMAND_EXECUTION]: The skill's operation depends on the execution of local Python scripts to aggregate data from different sources.\n - Evidence: The workflow requires running
scripts/search_web.py,scripts/search_china.py, andscripts/search_twitter.py.
Audit Metadata