The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process arbitrary natural language requests to scaffold and build applications, which creates a surface for indirect prompt injection attacks.
Ingestion points: Untrusted user input enters the agent's context through initial application requests (identified in SKILL.md and project-detection.md).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat user-provided app requirements as data rather than instructions, potentially allowing a user to inject malicious commands into the generation pipeline.
Capability inventory: The skill possesses extensive capabilities, including the Bash, Write, Edit, and Agent tools, which are used to execute shell commands and modify the file system across all project templates.
Sanitization: The instructions do not mention any sanitization, validation, or escaping of user input before it is utilized in shell commands or interpolated into generated source code.

app-builder