tavily-monitor
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPERSISTENCEEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external web searches and page extractions to compare current results against a saved baseline. \n
- Ingestion points: The agent reads
snapshot.jsonwhich contains URL snippets, page titles, and extracted content originating from the internet. \n - Boundary markers: No specific delimiters are defined to isolate untrusted data when it is read back into the agent's context for comparison. \n
- Capability inventory: The skill executes the
tvlytool and performs local file system operations (read/write) within a dedicatedmonitors/folder. \n - Sanitization: The skill does not prescribe specific sanitization or filtering of the external data before the agent processes it to detect changes. \n- [PERSISTENCE]: The skill provides instructions for establishing recurring checks on a cadence (e.g., daily or weekly). This involves using environment-specific scheduling tools to maintain the monitoring activity over time, which is only activated upon explicit user request. \n- [EXTERNAL_DOWNLOADS]: The documentation references the installation of additional search and research skills from the author's official repository using the platform's skill management command.
Audit Metadata