food_shopping-restaurants

The skill consists of three files: SKILL.md, metadata.json, and references/README.md. All files are purely descriptive markdown and JSON, providing instructions and metadata for the AI agent. There is no executable code, shell commands, or external resource fetching mentioned in any of the files. The skill's purpose is to act as a 'Restaurants advisor assistant' based on natural language instructions.

No prompt injection patterns were found. The descriptions are benign and do not attempt to override AI behavior. There are no commands for data exfiltration, such as reading sensitive files or making network requests. No obfuscation techniques (like Base64, zero-width characters, or homoglyphs) were detected. The skill does not instruct the installation of any external packages or the downloading of scripts, thus posing no unverifiable dependency risk. There are no privilege escalation attempts (e.g., 'sudo' commands) or persistence mechanisms. Metadata fields are clean and consistent with the skill's stated purpose. As the skill is purely instructional and does not process external user-provided content in a way that would introduce risk, indirect prompt injection is not a concern. Finally, no time-delayed or conditional attack patterns were identified. The skill is entirely declarative and relies on natural language processing without any underlying code execution.