cctv-news-fetcher-hardened
Fail
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The script at
scripts/news_crawler.jshardcodes a tracking/session cookie ('Cookie': 'cna=DLYSGBDthG4CAbRVCNxSxGT6') in the HTTP request headers. Hardcoding tokens or session identifiers is an insecure practice that can lead to credential exposure or session hijacking. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests external news content from the web for processing. 1. Ingestion points: Data is fetched from
cctv.cntv.cnandtv.cctv.comviascripts/news_crawler.js. 2. Boundary markers: The skill's instructions lack delimiters or isolation markers to distinguish between trusted instructions and untrusted crawled content. 3. Capability inventory: The skill is authorized to execute its crawler script usingbunornode. 4. Sanitization: The crawler performs minimal string cleanup (removing '[视频]' and trimming whitespace) but lacks robust sanitization to prevent the agent from interpreting instructions embedded in the news text.
Recommendations
- AI detected serious security threats
Audit Metadata