naver-shopping-search
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs the agent to execute code dynamically using
npx -y @nomadamas/k-skill@0. This command is used to fetch instructions, list files, and read the legal disclaimer from a remote package registry (NPM) at runtime. - [COMMAND_EXECUTION]: The skill instructions involve shell commands including
npxandcurlto interact with external services and regional repositories. - [EXTERNAL_DOWNLOADS]: The skill makes network requests to
https://k-skill-proxy.nomadamas.orgto perform product searches and retrieve retail data. These communications are directed to the vendor's infrastructure. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted third-party data from Naver Shopping (product titles, mall names, and metadata). This creates an attack surface where malicious content embedded in product listings could attempt to influence the agent's behavior.
- Ingestion points: External search results retrieved from the proxy endpoint in
instruction.md. - Boundary markers: None explicitly defined in the file contents to separate product data from instructions.
- Capability inventory: Perform network requests via curl, execute shell commands via npx, and surface product links to the user.
- Sanitization: The instructions recommend a "conservative" response style, but no technical sanitization logic is provided to filter the content of search results.
Audit Metadata