d6e-prompt-driven-ui
Pass
Audited by Gen Agent Trust Hub on Jul 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes platform tools like
d6e_sql,d6e_call_external_api, andd6e_update_workspace_prompt_ruleto fetch workspace data and modify agent instructions. The 'scenario-append' pattern specifically automates the update of prompt rules with workspace-specific values. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its processing of untrusted data from user inputs, Google Drive files, and SaaS APIs. Ingestion points include user-provided XML tags and Drive file content. Delimiters like XML tags and fenced code blocks are used as boundary markers. The skill possesses capabilities like SQL execution and prompt modification. Sanitization is addressed by recommending that markdown renderers escape HTML and restrict link schemes.
- [SAFE]: The skill implements security best practices, including same-origin download proxies to protect credentials and Zod for strict JSON schema validation.
Audit Metadata