The Agent Skills Directory

[PROMPT_INJECTION]: The SKILL.md file contains a meta-instruction: "IMPORTANT: When scripts fail, try to fix them directly." This instructs the AI agent to bypass failures by modifying its own code or the project environment, which could lead to the introduction of vulnerabilities or unintended behavior.
[COMMAND_EXECUTION]: The social photo design workflow in references/social-photos-design.md involves executing shell commands to launch Chrome in headless mode for taking screenshots, which requires spawning subprocesses using user-controlled parameters.
[EXTERNAL_DOWNLOADS]: The skill generates HTML files (presentations and social images) that include external assets from well-known services, such as Chart.js from cdn.jsdelivr.net and typography from fonts.googleapis.com.
[DATA_EXPOSURE]: The generation scripts in the scripts/ directory utilize a load_env function that attempts to harvest credentials from .env files located in the user's home directory (~/.claude/.env). This constitutes broad file system access for sensitive information retrieval.

design-pro