talk-dig
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it automatically fetches and processes external web content to generate reports.\n
- Ingestion points: Untrusted data enters the agent's context through
mcp__web_reader__webReaderwhich fetches full paper content from URLs discovered during the web search phase, as described inreferences/implementation.md.\n - Boundary markers: The implementation lacks explicit delimiters or instructions to the LLM to ignore potentially malicious instructions embedded within the fetched academic papers or abstracts.\n
- Capability inventory: The skill utilizes image analysis (
mcp__4_5v_mcp__analyze_image), web search (mcp__web-search-prime__webSearchPrime), and web content reading (mcp__web_reader__webReader). It does not have access to file-system writes or shell command execution.\n - Sanitization: The workflow does not include sanitization or filtering of the external text content before it is processed by the agent for extraction and summarization.
Audit Metadata