software-ai-integration
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The file
references/prompt-injection-and-ai-act.mdcontains examples of prompt injection attacks, including 'Ignore previous instructions' and 'DAN' (Do Anything Now) patterns. These are documented strictly for educational purposes as part of a security taxonomy to help developers recognize and defend against such attacks in their own integrations. - [EXTERNAL_DOWNLOADS]: The skill references documentation and code repositories from trusted organizations and well-known services (such as Vercel, Anthropic, OpenAI, and LangChain) in
data/sources.json. These links are provided as legitimate resources for product integration and do not involve untrusted third-party code execution. - [COMMAND_EXECUTION]: The provided utility script
scripts/check_injection_defenses.pyis a safe administrative tool. It performs static analysis of local JSON configuration files to ensure security best practices (like tool allowlisting and output validation) are implemented. It does not perform any network operations or unauthorized system modifications.
Audit Metadata