Make Custom Plugin Skill

Use this skill when asked to create or modify analysis plugins for MultiCellPose.

Goal

Produce a valid plugin Python file under guv_app/plugins/ that implements AnalysisPlugin in guv_app/plugins/interface.py, plus optional visualization and parameter schema.

Inputs You Should Ask For

1. •Plugin name (display name).
2. •Exact analysis objective.
3. •Required outputs (columns in result DataFrame).
4. •Hyperparameters with defaults/ranges.
5. •Whether visualization is required.

Required Interface Contract

Implement:

• •name property returning display string.
• •run(image, masks, classes=None, **kwargs) -> pandas.DataFrame.

Optional:

• •get_parameter_definitions() returns dict schema.
• •visualize(...) -> np.ndarray | None.

Implementation Workflow

1. •Inspect guv_app/plugins/interface.py.
2. •Mirror style from guv_app/plugins/basic_stats.py and guv_app/plugins/perimeter_intensity.py.
3. •Handle mask shape robustness:
   • •Prefer (H, W) masks.
   • •Support singleton 3D masks (1, H, W) when practical.
4. •Validate visualization masks with:
   • •from guv_app.plugins.validator import validate_visualization_mask
5. •Keep plugin side-effect free:
   • •Return a DataFrame, do not write files.
6. •Register by placement only:
   • •Place file in guv_app/plugins/; discovery is automatic.

Validation Step (Mandatory)

Run:

python scripts/validate_plugins.py

If failing, fix plugin and re-run.

Output Checklist

• •New plugin file in guv_app/plugins/.
• •Parameter definitions documented in code.
• •run() returns stable DataFrame schema.
• •visualize() returns valid mask or None.
• •Validator passes.