The Agent Skills Directory

[SAFE]: The skill is composed of documentation, rule sets, and character mappings for visual styling. It does not include any executable code or scripts that run on the host system, significantly limiting the attack surface.
[PROMPT_INJECTION]: While the instructions use imperative language to direct the model's output (e.g., "Render immediately", "do NOT call AskUserQuestion"), these directives are scoped to the tool's primary purpose of formatting and do not attempt to bypass safety filters or override core system instructions.
[DATA_EXPOSURE]: A thorough review of the skill's metadata and configuration files confirms that no hardcoded credentials, API keys, or sensitive environment variables are present. The skill only accesses the conversation context or user-provided topics for rendering.
[COMMAND_EXECUTION]: The skill does not utilize tools for arbitrary command execution. It only requests access to standard read tools (Read, Grep, Glob) to reference its own internal character and rule files, which is a benign and necessary operation for its functionality.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user input or conversation history to generate visual diagrams. While this constitutes a data ingestion surface, the skill lacks the high-risk capabilities (such as network access or shell execution) required to exploit potential injection content, and it does not interpolate this data into sensitive sinks.

quickviz