The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it fetches and processes untrusted content from external URLs without adequate safety boundaries.
Ingestion points: External technical articles are retrieved via the WebFetch tool based on user-provided URLs.
Boundary markers: There are no instructions to use delimiters or ignore embedded instructions within the fetched text, increasing the risk that the agent might obey malicious commands hidden in an article.
Capability inventory: The skill has file system write capabilities (mcp__filesystem__write_file, Write) and network read capabilities (WebFetch), which could be leveraged if an injection attack succeeds.
Sanitization: The instructions do not define any sanitization, filtering, or validation steps for the content retrieved from remote sources before it is processed.

translate-it-article