The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses instructional language like 'CRITICAL: Auto-start on load' and 'Crucial Style Instructions', but these are within the context of the skill's primary function and do not attempt to bypass AI safety filters or exfiltrate system prompts.
[DATA_EXFILTRATION]: The skill reads from local files such as about-me.md and project post files to personalize content. There are no network calls or attempts to send this data to external servers.
[COMMAND_EXECUTION]: The skill generates HTML files and text-based image prompts. It does not execute shell commands or system processes.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided post content to generate graphics. While it lacks explicit 'ignore embedded instructions' delimiters, the risk is minimal as the output is constrained to an HTML file or a structured prompt for the user to manually copy into an image generator. The capability tier is low.

graphic-designer