microscopy-particle-analysis

name	microscopy-particle-analysis
description	Counts and measures particles in a microscopy image (SEM, TEM, optical). Use when the user asks about particle count, particle size, grain size, size distribution, or wants to analyze blobs/spots in a 2-D micrograph image file.
when_to_use	When the user provides a path to a microscopy image (PNG, TIFF, JPG) and asks how many particles are in it, what their average size is, or how the size is distributed.
allowed-tools	Read Bash(python3 *) Bash(python *) Bash(pip install *)

Microscopy Particle Analysis

Task

Run the bundled analyze.py script on the user's microscopy image and report the results in plain language.

How to run

The script lives next to this SKILL.md. Invoke it with the user's image path. The script prints JSON to stdout.

python3 "${CLAUDE_PROJECT_DIR}/.claude/skills/microscopy-particle-analysis/analyze.py" <image_path> --min-area 30

If the user-skill version is installed at ~/.claude/skills/, use that path instead:

python3 "$HOME/.claude/skills/microscopy-particle-analysis/analyze.py" <image_path> --min-area 30

Optional flags

• --min-area N — minimum particle area in pixels (default 30). Increase to filter noise blobs, decrease to catch small particles.
• --save-overlay path.png — write a 3-panel diagnostic plot (original / threshold / labels) to the given path. Recommended whenever the user might want to verify the segmentation visually.

Dependencies

The script needs numpy, scipy, Pillow (or matplotlib). If a missing import error appears, install with:

pip install numpy scipy pillow matplotlib

How to report the result

Parse the JSON the script prints, then write a 2-3 sentence summary covering:

1. Particle count
2. Mean and standard deviation of equivalent radius (in pixels — note the user may need to convert via the image scale bar to physical units; if you know the scale, do the conversion explicitly)
3. The Otsu threshold value and the min_area setting used (so the result is reproducible)

If --save-overlay was used, end by mentioning the file path so the user can inspect the segmentation visually.

When to NOT call this skill

• The user is asking about a 1-D spectrum (XRD pattern, Raman, etc.) — wrong pipeline.
• The image is a structural/atomic-resolution micrograph where the goal is lattice analysis, not particle counting.
• The user has already segmented the image and just wants statistics — they can post-process directly.

In those cases, defer to the user or suggest a more appropriate workflow rather than forcing this script.