The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection through the processing of untrusted data.
Ingestion points: The debug_get_logs, debug_get_errors, and debug_get_stack_trace tools read directly from the Unity console, which often contains output from external scripts, network requests, or user-provided data.
Boundary markers: Absent. There are no instructions or delimiters used to prevent the agent from obeying commands embedded within log messages or stack traces.
Capability inventory: The skill has the ability to trigger script recompilation (debug_force_recompile) and modify global project configuration (debug_set_defines).
Sanitization: Absent. Data from the console is passed directly to the agent without filtering or escaping.
[COMMAND_EXECUTION]: Several tools perform high-impact operations within the Unity environment.
Evidence: debug_force_recompile triggers the Unity compilation engine, and debug_set_defines modifies persistent project-level scripting define symbols, which can alter the behavior of existing code gated by preprocessor directives.
[DATA_EXPOSURE]: The skill facilitates environment fingerprinting.
Evidence: debug_get_system_info and debug_get_stack_trace expose details about the host system environment, Unity version, and local file paths where the project is stored.

unity-debug