cp-prepare-segmentation-inputs¶
cp-prepare-segmentation-inputs resolves which image fields will enter segmentation.
It is the planning step before any CellProfiler segmentation run begins.
Purpose¶
Use this skill when you want:
confirm which fields will be sent into segmentation
inspect the segmentation input table before a longer run
hand a clean field list to cp-extract-segmentation-artifacts
Main Outcome¶
After this skill finishes, you know exactly which image fields will go into segmentation.
This is a planning step. It does not produce masks, outlines, or single-cell crops yet.
Inputs¶
This skill reads:
a project config such as
configs/project_config.demo.jsonthe configured raw-image and illumination roots
optional plate, well, or site filters
an optional output directory
Outputs¶
load_data_for_segmentation.csvThe field-level segmentation input table.pipeline_skill_manifest.jsonThe run record for this skill invocation.
Agent Demo¶
This page is based on a real local OpenClaw turn against the repository demo config:
session id:
segprepare-rerun-v2recorded on
2026-05-06model:
vibe/gpt-5-mini
Request¶
Before I segment the demo Cell Painting images, I want to confirm exactly which wells and image fields will be included in this run. Please use the demo project config, prepare the segmentation input table, write it under demo/workspace/outputs/agent_demo_segmentation/01_prepare_inputs_rerun_v2, and summarize the wells and fields that will be processed.
Structured Trace¶
user_input:
Before I segment the demo Cell Painting images, I want to confirm exactly which wells and image fields will be included in this run. Please use the demo project config, prepare the segmentation input table, write it under demo/workspace/outputs/agent_demo_segmentation/01_prepare_inputs_rerun_v2, and summarize the wells and fields that will be processed.
recorded_turn:
- the saved OpenClaw turn answered this request directly
- the reply identified the segmentation-input step used
- the reply reported the output path of the load-data table
- the reply summarized the number of wells and fields included
- the reply listed the exact well/site combinations
final_answer_snapshot:
The demo segmentation run includes two image fields: BR00000001 / A01 / site 1 and BR00000001 / A02 / site 1.
Observed Result¶
The recorded run wrote:
load_data_for_segmentation.csvpipeline_skill_manifest.json
The resolved segmentation scope covered:
1plate2wells2image fields
The included fields were:
BR00000001 / A01 / site 1BR00000001 / A02 / site 1
Recorded Reply¶
Done — I used the stable step **`cp-prepare-segmentation-inputs`** with the demo project config and wrote the segmentation input table under:
- `demo/workspace/outputs/agent_demo_segmentation/01_prepare_inputs_rerun_v2`
### Output file
- `demo/workspace/outputs/agent_demo_segmentation/01_prepare_inputs_rerun_v2/load_data_for_segmentation.csv`
### What will be processed
This prepared segmentation run includes:
- **1 plate**
- **2 wells**
- **2 image fields (rows in the load-data table)**