The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and retrieval of document content, which introduces a potential attack surface for indirect prompt injection if the stored content contains malicious instructions.\n- Ingestion points: External data enters the system via collection.add() as shown in SKILL.md and through framework-specific loaders like Chroma.from_documents() in both SKILL.md and references/integration.md.\n- Boundary markers: The implementation examples do not demonstrate the use of delimiters or specific prompt engineering to isolate retrieved data from system instructions.\n- Capability inventory: The skill allows for local file system operations through chromadb.PersistentClient and network requests via chromadb.HttpClient.\n- Sanitization: The provided code snippets do not include logic for sanitizing or validating ingested document strings before they are processed by the vector database or returned to the agent context.

chroma