The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill analyzes external mobile codebases and possesses high-privilege tools, creating a vulnerability surface where malicious instructions in project files could influence agent behavior. 1. Ingestion points: The skill uses read_file and edit_file to process source code from mobile projects. 2. Boundary markers: The skill does not define delimiters or specific instructions to the agent to ignore or isolate instructions found within those files. 3. Capability inventory: The skill is equipped with run_command, write_file, edit_file, and read_file tools. 4. Sanitization: There are no mechanisms for validating, escaping, or sanitizing the content ingested from external sources before it is processed by the model.

resonance-mobile