x-twitter-browser-ops
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from X/Twitter which could potentially contain adversarial instructions designed to influence the agent's behavior. * Ingestion points: Extracts user profiles, post content, and relationship data from X/Twitter browser pages and GraphQL responses (SKILL.md). * Boundary markers: Instructions do not mandate specific delimiters to isolate external text within the agent's context. * Capability inventory: Includes browser navigation, bookmark mutations, and file generation (CSV, MD, HTML). * Sanitization: The skill explicitly directs the agent to escape JSON characters when generating HTML dashboards to prevent data-driven injection in the resulting artifacts.
- [DYNAMIC_EXECUTION]: The skill contains instructions for generating PowerShell scripts to build HTML visualization dashboards. * Evidence: Detailed guidance on using PowerShell to compress JSON and escape characters for HTML embedding (SKILL.md). * Context: The execution is limited to generating static reports from collected data, with built-in safety recommendations for character escaping.
Audit Metadata