The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and learn from external Kaggle competition URLs. This creates a potential surface where untrusted content from those URLs could influence the agent's behavior if the data is processed without strict sanitization.
[DYNAMIC_EXECUTION]: Multiple reference files (e.g., nlp/aimo-2-2025.md) contain educational code snippets demonstrating the use of exec() and subprocess.run(). While these are documented as templates for 'safe code execution' in sandboxed environments for mathematical reasoning or automated bug fixing, they represent powerful capabilities that require careful implementation in production environments.
[EXTERNAL_DOWNLOADS]: The skill's documentation includes templates for web scraping (using requests and BeautifulSoup) and calls to external APIs (e.g., openai). These are directed at legitimate domains such as Kaggle, arXiv, and Art of Problem Solving for data retrieval purposes.
[SAFE]: No obfuscation, persistence mechanisms, or hardcoded credentials were detected. The skill primarily consists of technical documentation and educational code templates for machine learning practitioners.

kaggle-learner