The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill ingests untrusted data that could influence the agent's context through its output.\n
Ingestion points: The scripts scripts/scaffold_app.py and scripts/scaffold_app.js accept the project_type as a command-line argument from the user.\n
Boundary markers: Absent. The scripts do not use any delimiters to separate user input from system-generated logs.\n
Capability inventory: All scripts in the current version are limited to print and console.log operations. There are no file system modifications, network requests, or subprocess executions implemented in the provided snippets.\n
Sanitization: No validation or escaping is performed on the project_type input before it is printed to the terminal, where an agent might read and interpret it.\n- [Remote Code Execution] (SAFE): The execution protocol references local scripts within the skill directory. No patterns for downloading and executing remote code (e.g., piping curl to bash) were identified.\n- [Data Exposure & Exfiltration] (SAFE): No hardcoded credentials, access to sensitive file paths (like ~/.ssh), or outbound network connections were detected.\n- [Privilege Escalation] (SAFE): No use of sudo, chmod, or other privilege-altering commands were found in the scaffolding instructions or scripts.

full-stack-scaffold