The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface by processing instructions and code from an external platform.
Ingestion points: The prompt field in theme definitions (accessed via magicpath-ai get-theme) and the component source code (accessed via magicpath-ai inspect) provide natural-language instructions and executable code from the platform to the agent.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the processing workflow.
Capability inventory: The skill utilizes the magicpath-ai CLI for shell execution, can write files to the local repository via the add command, and is instructed to modify existing application code for integration.
Sanitization: No sanitization, validation, or escaping of the platform-provided data is defined.
[COMMAND_EXECUTION]: The skill uses shell commands for environment management and platform interaction.
Employs dynamic context injection (!) in SKILL.md to automatically execute magicpath-ai info when the skill is loaded to determine auth and project status.
Directs the agent to perform global installation of the magicpath-ai package via npm and manages local filesystem writes using the CLI's installation capabilities.

magicpath