Guide · Agent skill
How to use the CellProfiler Nuclei Assay agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The CellProfiler Nuclei Assay skill gives an AI agent a standardized way to run CellProfiler on fluorescence microscopy images. It finds and counts cell nuclei in 2D grayscale images and measures each nucleus's DNA stain intensity, area and shape. This guide explains what the skill does, how to install it, how to ask for it in plain language, and what its outputs do and do not prove.
CellProfiler Nuclei AssayK-Dense-AI/scientific-agent-skills · MIT Download the skill Quick start
- 1Download and unzip. Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder.
- 2Check your images. Confirm each image is a single-channel 2D grayscale TIFF in uint8 or uint16 format, with no RGB, z-stacks or float data.
- 3Make the manifest. Create images.csv with a unique, nonnumeric sample ID, the image path, and plate, well and site fields.
- 4Run a pilot and check overlays. Run a small pilot covering controls, dense and dim fields, and plate edges, then inspect the saved overlays for missed nuclei, merges and splits.
- 5Run the batch. Run prepare, then run into a fresh output folder, then summarize the results.
- 6Review the QC file. Open assay_qc.json and cellprofiler.log to confirm the run completed and the checks passed.
What the skill makes the agent do
An agent skill is a folder of instructions and helper scripts that an AI agent can load when a task matches it. This one covers repeatable CellProfiler microscopy pipelines for nuclear segmentation, cell counts and per-object fluorescence measurements on 2D fluorescence assays.
The agent checks the image manifest and input files, runs CellProfiler in batch mode, records checksums and run status in a QC file, and summarizes the resulting tables. It does not replace visual inspection of the segmentation overlays, which the skill itself says are required.
- Creates a CSV manifest of images and a load_data file for CellProfiler
- Runs the bundled starter pipeline in headless batch mode with a saved copy of the pipeline
- Writes assay_qc.json with run status, checksums, and per-sample checks
- Summarizes the output tables once a run has finished
Install the skill from the ZIP
Download the skill ZIP from this page and unzip it. The skill folder contains SKILL.md, the scripts/nuclei_assay.py helper, and references/runtime-and-qc.md.
For a personal install, unzip the folder into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. Either way, the skill folder should sit directly inside the skills folder.
You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
Check the requirements before you start
The skill needs Python 3.12 or newer with numpy and tifffile for its helper scripts. Segmentation needs a separate CellProfiler application or a local container launcher. The skill targets a specific CellProfiler release, which its runtime reference describes in detail.
Your images must be single-channel 2D TIFFs in uint8 or uint16 format. The helper rejects RGB images, z-stacks and float images rather than guessing what they contain. The skill needs no credentials.
- Python 3.12+ with numpy and tifffile
- CellProfiler application or local container launcher for segmentation
- Input: 2D grayscale uint8 or uint16 TIFF, one nuclear channel per field
- Network access is needed only for installation
Safety notes: scripts and network
This skill runs scripts. The helper script and the CellProfiler executable it calls will run on your machine, so review the files before you install and run the skill on data you care about.
The skill does not use the network during runs. Network access is needed only to install the software it depends on.
- Runs scripts: yes
- Uses network: no
- Keep original image intensities; convert proprietary formats explicitly before using the helper
Ask for it with natural requests
You do not need to name the skill or any commands. Describe the assay and the result you want, and the agent will load the skill when the request matches. Good requests name the image type, the nuclear channel and the measurement you need.
Examples of requests that should trigger the skill:
- Count the nuclei in these DAPI TIFF fields and export per-nucleus DNA intensity.
- Set up a reproducible CellProfiler batch run for my control and treated wells.
- Make a manifest for these single-channel fluorescence images and check it before running.
- Summarize the Nuclei.csv and Image.csv from my last CellProfiler run and flag problems.
Run the assay from the command line
The agent normally runs these steps for you. You can also run them yourself from the skill folder. First create images.csv listing each image, then prepare, run and summarize.
The run step requires a fresh, empty output directory. Rerun in a new output folder rather than reusing an old one.
- prepare builds the load_data file and works without CellProfiler
- run executes CellProfiler and needs the executable named with --executable
- summarize checks the CSV contents, but does not prove that an engine run completed
python scripts/nuclei_assay.py prepare images.csv load_data.csv
python scripts/nuclei_assay.py run images.csv results --executable cellprofiler
python scripts/nuclei_assay.py summarize resultsRead the outputs and QC files
Each run writes results you can check. Start with the overlay images, because a correct count is not enough to show that the nuclei were segmented correctly. Intensities are scaled by the integer storage maximum, so compare them only within one bit depth, exposure and staining batch.
A saturated image can pass segmentation while its intensity measurement is unusable. Illumination correction and background subtraction are not applied by this starter pipeline.
- Image.csv: one row per image or field, including Count_Nuclei
- Nuclei.csv: one accepted nucleus per row, with intensity, area and shape
- *_nuclei.png: green nuclear boundaries over the input for visual QC
- pipeline.cppipe and cellprofiler.done: the exact pipeline copy and completion marker
- assay_qc.json: run status, checksums, counts, intensity and area checks, and saturation flags
- cellprofiler.log: CellProfiler's output
What people use it for
- Counting nuclei per field in DAPI-stained single-channel fluorescence images
- Comparing per-nucleus DNA stain intensity between control and treated wells, with calibrated images
- Producing a documented, repeatable CellProfiler batch run with a saved pipeline and checksums
- Checking a batch for saturated images, zero counts and unusual count or area distributions
Frequently asked questions
Is the CellProfiler Nuclei Assay skill free to use?
The skill is released under the MIT licence. CellProfiler itself is a separate application, and you need it, or a local container launcher, for segmentation.
What do I need installed to use this skill?
You need Python 3.12 or newer with numpy and tifffile for the helper scripts. You also need a CellProfiler application or a local container launcher, and your images must be 2D grayscale uint8 or uint16 TIFFs.
Can the skill analyse z-stacks or multichannel cell painting images?
No. The skill is built for one 2D grayscale nuclear channel per field. Volumetric, multichannel cell painting and tissue-specific segmentation need a separate, validated pipeline, and forcing this one on them gives wrong results.
Why does the helper reject my TIFF files?
It rejects RGB images, z-stacks and float images rather than guessing channels. Palette and white-is-zero TIFFs need an explicit conversion first.
Can I use this for clinical or diagnostic decisions?
No. The skill checks counts and overlays but does not prove that segmentation is correct, so it is not suitable for clinical or diagnostic decisions.
How does this compare with running CellProfiler by hand?
The skill sets up a standardized pipeline and records checksums, run status and sample QC automatically. Running CellProfiler by hand gives you full control over the pipeline, which you will still need for assays the skill does not cover.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you run nuclear counts or nuclear DNA intensity measurements on single-channel 2D TIFF images and want a reproducible, documented CellProfiler batch run. Pick something else if your images are volumetric, multichannel or tissue-based, or if you need clinical results, because the skill cannot confirm that segmentation is correct.
Download the skill