customer-insights
Pass
Audited by Gen Agent Trust Hub on Feb 18, 2026
Risk Level: SAFENO_CODE
Full Analysis
- PROMPT_INJECTION (SAFE): No patterns of instruction override, jailbreaking, or system prompt extraction were found. The instructional language is benign and follows the intended role-play for customer analysis.
- DATA_EXFILTRATION (SAFE): No hardcoded credentials, API keys, or access to sensitive file paths (e.g., .ssh, .aws) were detected. There are no network operations (curl, fetch) to external domains.
- REMOTE_CODE_EXECUTION (SAFE): There are no scripts, package manager files (package.json, requirements.txt), or commands that download and execute external content.
- INDIRECT_PROMPT_INJECTION (SAFE): While the skill is designed to process untrusted customer data provided by the user, it lacks any dangerous capabilities such as file system writes, network requests, or shell execution, neutralizing the risk of indirect injection.
- OBFUSCATION (SAFE): The content is entirely in plain text with no Base64, zero-width characters, or hex-encoded commands.
Audit Metadata