The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill documents administrative functionality in references/advanced.md and references/connection.md, specifically the suConn method. This capability allows an authenticated administrator to create a connection acting as any other user on the system, which enables user impersonation and broad data access beyond the initial account's scope.
[DYNAMIC_EXECUTION]: The references/scripts.md file provides a detailed guide for creating and deploying OMERO.scripts. These are Python scripts that are uploaded to and executed on the OMERO server, providing a mechanism for server-side code execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-controlled data retrieved from remote OMERO servers, which can influence the agent's behavior.
Ingestion points: Metadata retrieval methods such as image.getDescription() (data_access.md), ann.getValue() (metadata.md), and ROI labels (rois.md).
Boundary markers: Absent; data is processed and potentially interpolated into the agent's context without delimiters or warnings.
Capability inventory: The skill enables file system writes (open(file_path, 'wb')), server-side script deployment, and administrative actions.
Sanitization: None detected; the skill assumes all metadata retrieved from the OMERO server is safe to process.
[CREDENTIALS_UNSAFE]: The documentation in references/advanced.md and references/connection.md includes examples with hardcoded credential placeholders (e.g., ADMIN_PASS = 'password', PASSWORD = 'pass'). While intended as examples, they demonstrate a pattern of handling sensitive authentication data in plain text.
[COMMAND_EXECUTION]: The skill provides shell commands for environment preparation and workflow management, including package installation (uv pip install omero-py) and CLI-based script management (omero script upload, omero script launch).

omero-integration