The Agent Skills Directory

[DATA_EXPOSURE]: The skill follows security best practices for credential management by utilizing environment variables (e.g., os.environ.get('VOYAGE_API_KEY')) in its code templates rather than hardcoding sensitive information.
[DYNAMIC_EXECUTION]: In references/langchain-architecture.md, the skill implements a math calculation tool using the ast module to safely parse expressions. This is a robust security measure that prevents arbitrary code execution vulnerabilities commonly associated with the eval() function.
[INDIRECT_PROMPT_INJECTION]: While the skill's primary purpose involves building RAG and agent systems that ingest untrusted data, it proactively includes templates for validation, verification, and groundedness checks (e.g., in references/llm-evaluation.md and references/prompt-engineering-patterns.md) to mitigate injection risks.

ai-engineer