technology-search
Pass
Audited by Gen Agent Trust Hub on Apr 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/search_news.jsutilizesexecFileto invoke a shell script from the siblingweb-searchskill (web-search/scripts/search.sh) as a fallback mechanism for broader results. - [EXTERNAL_DOWNLOADS]: The skill performs multiple outbound HTTP requests to fetch RSS feeds and API data from a curated list of tech news providers, including Hacker News (via Algolia API), Reddit, and various official technology blogs.
- [DATA_EXFILTRATION]: The
scripts/shared/network_detector.jsmodule performs connectivity tests against Cloudflare and TechCrunch to determine network reachability and automatically adjust the news source region, which is a standard functional requirement for this skill. - [PROMPT_INJECTION]: The skill processes untrusted third-party data (titles and summaries) from external RSS feeds, creating an indirect prompt injection surface.
- Ingestion points: Third-party data is ingested via
scripts/parsers/rss_parser.jsandscripts/parsers/hn_parser.js. - Boundary markers: The output returned to the agent does not use explicit boundary markers to wrap content from external feeds.
- Capability inventory: The skill has network access and subprocess execution capabilities.
- Sanitization: The
cleanTextfunction inscripts/parsers/rss_parser.jsremoves HTML tags and decodes common entities, providing basic sanitization of the external content.
Audit Metadata