The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes microscopy metadata, annotations, and table data retrieved from an external OMERO server. These fields could contain malicious instructions designed to influence the agent. Given that the skill has the capability to delete objects and execute server-side scripts, it presents a functional surface for indirect prompt injection attacks. \n * Ingestion points: references/metadata.md (listAnnotations), references/data_access.md (object attribute retrieval), and references/tables.md (table data reading). \n * Boundary markers: No explicit delimiters or instruction-ignoring warnings are present in the provided templates for processing external data. \n * Capability inventory: Object deletion (references/advanced.md), local file system writes for downloads (references/metadata.md, references/advanced.md), and deployment of server-side Python scripts (references/scripts.md). \n * Sanitization: The skill uses official omero.gateway wrappers which provide structured data access, but it does not perform specific sanitization of the natural language content within metadata fields before processing. \n
[SAFE]: The skill implements best practices for authentication, recommending the use of environment variables and configuration files for credentials rather than hardcoding. It also correctly uses yaml.safe_load for configuration parsing.

omero-integration