watch-for-changes

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPERSISTENCEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on the execution of tvly and python3 commands to extract web content and perform comparison logic.
  • [EXTERNAL_DOWNLOADS]: The documentation points users to the official Tavily CLI GitHub repository for installation guidelines.
  • [PERSISTENCE]: The skill instructs the agent to establish recurring execution using standard operating system tools such as cron jobs, systemd timers, or Windows Task Scheduler to maintain the monitoring schedule.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the web, which represents a potential injection surface.
  • Ingestion points: Raw content is fetched from external URLs via tvly extract (SKILL.md).
  • Boundary markers: The scripts do not use specific delimiters or instructions to ignore embedded commands in the fetched content.
  • Capability inventory: The skill has access to python3, tvly (network), and file writing for snapshots (SKILL.md).
  • Sanitization: Content is hashed for comparison purposes, but raw text is stored and processed without sanitization.
  • [DYNAMIC_EXECUTION]: The skill uses inline Python script generation (python3 << 'PYEOF') to implement its snapshotting and diffing logic at runtime.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 06:49 PM
Security Audit — agent-trust-hub — watch-for-changes