The Agent Skills Directory

Category 8: Indirect Prompt Injection (LOW): The skill takes untrusted user input (meme text) to construct URLs. While this represents an ingestion point, the output is restricted to URL construction for a specific API.
Ingestion points: User-provided top_text, bottom_text, and template parameters in meme_generator.py and SKILL.md.
Boundary markers: The skill relies on standard URL encoding and custom character mapping (e.g., to _) to sanitize input for the memegen.link API.
Capability inventory: The skill constructs strings for display as markdown images. It contains no subprocess calls, file-write operations, or active network requests within its scripts.
Sanitization: Uses urllib.parse.quote and a structured mapping in _format_text to escape reserved characters.
Category 4: Unverifiable Dependencies & Remote Code Execution (SAFE): The provided Python script meme_generator.py uses only standard libraries (argparse, urllib.parse). No external packages are required or installed.
Category 2: Data Exposure & Exfiltration (SAFE): No access to sensitive file paths (~/.aws, ~/.ssh, etc.) or environment variables was detected. The skill only interacts with the public memegen.link API domain.

meme-factory