The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill ingests untrusted data from the TouchDesigner environment via tools like td_operators and td_selection, and it has the capability to execute arbitrary code via td_execute. There are no boundary markers or sanitization logic defined in the instructions to prevent data read from the environment (like operator names or DAT content) from being interpreted as instructions.
[Command Execution] (MEDIUM): The td_execute tool provides direct access to the Python interpreter within the host application. While necessary for the skill's intended functionality, it represents a high-impact execution vector that could be exploited if the agent is manipulated via prompt injection.
[Instructional Override] (LOW): The skill uses 'CRITICAL' and 'REQUIRED' markers to override the agent's pre-trained knowledge. While this is likely intended for technical accuracy, such self-referential instructions are a known pattern for steering agent behavior and bypassing general guidelines.

touchdesigner