The Agent Skills Directory

[PROMPT_INJECTION] (LOW): The skill explicitly instructs the agent to set 'safety_tolerance' to '6' to achieve 'maximum creative freedom,' which is a known pattern for bypassing built-in AI safety filters and content guidelines.
[COMMAND_EXECUTION] (MEDIUM): Multiple files in the skill definition use relative path traversal ('../../../') to reference executable scripts and configuration files. This attempts to force the agent to access or execute files outside of the skill's designated sandbox environment.
[CREDENTIALS_UNSAFE] (LOW): The skill manages sensitive API keys and references external credential files through path traversal, creating a potential vector for credential exposure if the execution environment is not strictly isolated.

image-generator