daily-ai-news
Fail
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: CRITICALEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill recommends the domain 'ai.hubtoday.app' as a 'Primary Source' (Tier 1) in both 'SKILL.md' and the 'references/news_sources.md' database. This specific domain has been flagged as malicious by automated scanners. By including this as a recommended resource, the skill encourages the AI agent to fetch content from a risky source and provides the end-user with potentially dangerous links in the final news briefing, which could lead to phishing or malware exposure.
- [INDIRECT_PROMPT_INJECTION]: The core workflow involves the automated ingestion of external web content using the 'mcp__web_reader__webReader' tool to create news summaries. This design presents a high-exposure surface for indirect prompt injection attacks. * Ingestion points: The agent fetches full article text from URLs provided in the source database and dynamic web search results, specifically targeting news sites including the blacklisted domain mentioned above. * Boundary markers: The skill lacks explicit delimiters or instructions for the agent to isolate the fetched content from its own internal logic, making it susceptible to instructions embedded within news articles. * Capability inventory: The agent has access to 'WebSearch' and 'mcp__web_reader__webReader'. While it lacks direct file-system write or shell capabilities, it could be manipulated into generating biased or harmful output based on malicious news content. * Sanitization: There are no safeguards described to filter, escape, or validate the Markdown content fetched from external sources before it is processed for summarization.
Recommendations
- CRITICAL: 2 file(s) identified as malware by FileRep - DO NOT USE
- AI detected serious security threats
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata