The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to crawl external web data (Google Image search results) including image titles and source URLs. This creates a surface for indirect prompt injection where malicious content on a webpage could influence the agent's behavior if it processes the crawler's output.
Ingestion points: scripts/crawl.py extracts data from web search results.
Boundary markers: None; search results are returned as structured JSON but without delimiters to warn downstream LLM processing.
Capability inventory: Network access for crawling and file system writes for saving results.
Sanitization: The provided scripts do not show sanitization of retrieved titles or metadata.
[Unverifiable Dependencies] (SAFE): The core logic for the crawler (core/crawler.py) is missing from the provided files. However, the interface in scripts/crawl.py and the configuration management in core/config.py do not exhibit malicious patterns or suspicious execution methods like eval or subprocess calls on untrusted data.

google-image-crawler