The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process data from external sources, such as remote URLs and local files, which is inherent to document analysis. If documents contain adversarial text, it could potentially influence the agent's behavior in downstream tasks.
Ingestion points: The client.beginAnalyzeDocument method in SKILL.md and references/examples.md processes data from user-provided URLs or local file paths.
Boundary markers: The provided Java snippets do not demonstrate specific delimiters or instructions to ignore embedded commands within the analyzed document content.
Capability inventory: The skill uses the Azure SDK to perform network requests to cloud services and file system operations to read document data.
Sanitization: Extracted text and fields are processed as raw data strings without explicit sanitization in the example code.
External Service Communication: The skill performs network operations to reach Azure AI services. These connections use official SDKs and secure authentication methods, which is a standard requirement for this functionality.
Local File Access: The code examples include patterns for reading local files (BinaryData.fromFile). While necessary for analyzing local documents, users should ensure the agent's file access is scoped to intended directories.

azure-ai-formrecognizer-java