ad-angle-research
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content retrieved from external ad libraries. It incorporates a defensive instruction for the AI to treat all scraped text (ad copy, landing pages, tool results) strictly as data and specifically ignore any commands embedded within that content, such as 'ignore previous instructions'. This is an industry-standard best practice for skills that handle external data.
- [EXTERNAL_DOWNLOADS]: The skill utilizes a third-party Model Context Protocol (MCP) connector from Tracetify to fetch live ad data. It provides the user with setup instructions to configure this tool, including a bearer token placeholder. The skill does not install the software automatically and requires the user to explicitly confirm before initiating lookups that consume service credits.
- [COMMAND_EXECUTION]: The skill documentation includes a setup command for the user to run in their terminal to enable the MCP connector. This command is provided for manual setup by the user and is not executed automatically by the skill.
Audit Metadata