gmgn-wallet-style
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted third-party data from the blockchain via
gmgn-clicommands, including token names and metadata labels (common.tags). This data is processed in the agent context, creating a potential surface for indirect prompt injection. The skill mitigates this by instructing the agent to treat such fields strictly as data and ignore any text resembling commands or overrides. - [DYNAMIC_EXECUTION]: The skill generates interactive HTML pages for visual previews (Step 9). This involves writing local HTML files with embedded JavaScript logic for UI scaling and tooltips. The skill includes specific sanitization requirements, such as HTML-escaping special characters and using unicode escapes for JSON embedded in scripts, to prevent cross-site scripting (XSS) or further injection.
- [COMMAND_EXECUTION]: The skill relies on shell command execution via the
gmgn-clitool. It implements defensive measures by requiring strict regex validation for wallet addresses and utilizing an allowlist for supported blockchain networks to prevent command injection vulnerabilities.
Audit Metadata