product-box-converter
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill fetches content from external websites via user-provided URLs using curl or WebFetch. This creates a surface for indirect prompt injection, where malicious content on the fetched page could attempt to influence the agent's behavior. Ingestion points: Path 3 (URL Input) in SKILL.md fetches untrusted HTML from user-supplied URLs. Boundary markers: No specific delimiters or instructions are provided to the agent to treat fetched content as data rather than instructions. Capability inventory: The skill instructions allow the agent to execute shell commands (pbcopy) and perform network requests (curl, WebFetch). Sanitization: While the skill suggests cleaning HTML entities and clutter, it lacks robust sanitization to prevent instruction injection from the scraped content.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute the pbcopy command using a heredoc to transfer results to the system clipboard. The use of an unquoted heredoc delimiter may allow the shell to interpret special characters if they are present in the scraped product data, potentially leading to unintended command execution.
- [EXTERNAL_DOWNLOADS]: The skill relies on curl and WebFetch to download product page content from arbitrary external URLs provided by the user.
Audit Metadata