The Agent Skills Directory

Prompt Injection (SAFE): The skill contains strict instructional logic, such as the blocking operations defined in first-time-setup.md, to manage the agent's workflow state. These are functional constraints designed to ensure proper configuration and do not attempt to bypass core AI safety filters or exfiltrate system prompts.
Data Exposure & Exfiltration (SAFE): No hardcoded credentials, API keys, or sensitive file paths (like SSH keys or AWS configs) were found. The skill does not perform unauthorized network requests; the provided script merge-to-pdf.ts only performs local file system operations.
Unverifiable Dependencies (SAFE): The PDF generation script uses the standard pdf-lib library. There are no patterns of downloading and executing remote scripts or unverified third-party packages at runtime.
Indirect Prompt Injection (LOW): The skill processes user-provided source material to generate image prompts. While this is a standard capability, there is a theoretical surface for indirect injection where source content could attempt to influence the visual output or instructions. The skill uses boundary markers (triple backticks) in its base-prompt.md template to help mitigate this risk.
Command Execution (SAFE): The skill includes a TypeScript script (merge-to-pdf.ts) intended for local execution via Bun. This script is limited to reading image files and merging them into a PDF document; it does not utilize eval(), exec(), or other dangerous dynamic execution patterns.

baoyu-comic