The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides several standalone Python scripts (e.g., inventory.py, export_image_metadata.py, plan_transfer.py) designed to be executed by the agent to automate microscopy workflows and validate configurations.
These scripts are localized to the skill's specific research purpose and include strict argparse validation to prevent arbitrary argument injection.
File operations within the scripts use atomic write patterns and enforce owner-only filesystem permissions for generated outputs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted metadata (such as image annotations, ROI labels, and tags) from remote OMERO servers, which is a potential injection surface.
Ingestion points: Metadata is retrieved through the omero-py gateway using methods like listAnnotations() and copyShapes() (referenced in references/metadata.md and references/rois.md).
Boundary markers: The instructions throughout the skill and its references emphasize the use of hard caps on result counts and string lengths to limit the impact of large or malicious payloads.
Capability inventory: The skill possesses the capability to perform network requests to OMERO servers, write JSON files to the local disk, and perform directory scans.
Sanitization: The skill includes a dedicated utility, omero_common.py, which provides a json_safe function used by all scripts to truncate long strings and limit the size of collections before they are processed by the agent.
[EXTERNAL_DOWNLOADS]: The skill guides the user to install omero-py and platform-specific zeroc_ice wheels necessary for Ice middleware communication.
These resources are sourced from the official OpenMicroscopy Environment (OME) organization and their documented commercial partner, Glencoe Software, which are established and trusted entities in the scientific community.

omero-integration