The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents how to build systems that ingest external documents into vector databases for RAG workflows. This pattern inherently creates a surface for indirect prompt injection if the ingested content is not sanitized before retrieval and processing by an LLM.
Ingestion points: The EmbeddingStoreIngestor patterns in SKILL.md and references/examples.md describe processing external document content.
Capability inventory: The skill enables database write capabilities via EmbeddingStore.add and addAll methods.
Boundary markers: No specific delimiters or validation logic for untrusted content are included in these configuration-focused snippets.
Sanitization: Input sanitization for ingested data is not addressed in this configuration guide.
[CREDENTIALS_UNSAFE]: The code snippets include standard documentation placeholders for database credentials (e.g., username, password). The skill correctly demonstrates best practices elsewhere by suggesting the use of environment variables for sensitive keys.

langchain4j-vector-stores-configuration