The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructions and code examples reference sensitive local configuration files for authentication.
Evidence: Mentions ~/.aws/credentials in the setup instructions and as a source for AWS authentication profiles.
[INDIRECT_PROMPT_INJECTION]: The skill provides capabilities to ingest and process untrusted data from external sources through Bedrock Knowledge Bases and tool-using agents, creating an attack surface for indirect prompt injection.
Ingestion points: AmazonKnowledgeBasesRetriever fetches data from external AWS Knowledge Bases; BrowserToolkit retrieves content from the web.
Boundary markers: None specified in the provided code examples or instructions to delimit retrieved content from system instructions.
Capability inventory: The skill uses ChatBedrockConverse and create_react_agent which can execute tools and generate responses based on the retrieved data.
Sanitization: No explicit sanitization or validation of the retrieved external content is described before it is processed by the LLM.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the official langchain-aws package and references documentation from external domains.
Evidence: Recommends pip install -U langchain-aws and links to github.com/langchain-ai/langchain-aws and reference.langchain.com.

langchain-aws