The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides boilerplate code for interacting with the host file system, which poses a risk if an agent implements these features and processes user-controlled paths without rigorous validation.
Ingestion points: The read_file and write_file Rust commands in SKILL.md take string arguments for paths and content.
Boundary markers: The code examples are generic templates and do not include specific boundary enforcement or validation logic, although the text guidelines mention them.
Capability inventory: Provides templates for full file read and write access using the Rust std::fs module.
Sanitization: The skill contains a 'Security' section advising developers to 'Validate all inputs from the frontend' and 'Sanitize file paths', mitigating the risk but leaving implementation to the agent/developer.
[COMMAND_EXECUTION]: The guidelines describe the creation of Tauri commands that allow the frontend (TypeScript) to execute Rust functions. This IPC mechanism is the primary way the skill facilitates interaction with host system resources and logic execution.

tauri-development