What does this skill do, and when should you use it?
image-crop-rotate is a single skill within the instavm/open-skills collection that performs two fixed operations on images using a Python PIL/Pillow script: center-cropping to 50% of original size, then rotating 90 degrees clockwise. The skill consists of a SKILL.md file and a scripts/crop_and_rotate.py script. It is designed to run inside OpenSkills' sandboxed container on your Mac, so files never leave your machine. The operations are fixed — custom crop ratios or rotation angles are not supported.
Reads an input image file and runs the crop_and_rotate.py script, which first crops to 50% of the original size from the center (e.g., 1000x800 becomes 500x400), then rotates the cropped image 90 degrees clockwise (500x400 becomes 400x500), and saves the result to the given output path. It supports any format PIL/Pillow handles (JPEG, PNG, GIF, BMP, TIFF, etc.) and prints original, cropped, and final dimensions.
- Users with sensitive screenshots or photos who want processing done locally without cloud uploads
- Organizers who need images halved in size and reoriented (one call processes one image)
- Users who already run the OpenSkills MCP server and want Claude Desktop or Gemini CLI to invoke image tools
- Developers verifying output geometry, since the script prints sizes at each stage
How do you install this skill?
- Static review only; no code was executed and all conclusions are from source reading.
- Hardcoded /mnt/user-data paths make the skill usable mainly inside the companion sandbox/MCP environment; adjust paths for standalone local use.
- EXIF orientation is not handled, so JPEGs with rotation metadata may produce unexpected results.
- Crop ratio and rotation angle are fixed at 50% / 90° clockwise; other needs require editing the script.
- Publisher is not verified by the FollowSkills registry; the skill has no independent version or changelog, so updates must be tracked in the repo.
- Pillow is not declared in the skill directory; ensure it is installed in the runtime environment.
- Shell / CLI
- Local filesystem
PythonPIL/Pillow
Prerequisites: OpenSkills installed per its README (Apple Silicon Mac, Python 3.10+, run ./install.sh and pip install -r examples/requirements.txt). The skill lives in the repo at skills/public/image-crop-rotate/. Whether install.sh automatically places it into ~/.open-skills/assets/skills/ is not documented in the source; verify after installation.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/instavm/open-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/public/image-crop-rotate" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Crop this image to 50% from the center and rotate it 90 degrees clockwise.
- Put your image in ~/.open-skills/assets/outputs/ (mapped to /mnt/user-data/ inside the container). 2. Ask your AI assistant, e.g., "Crop this image to 50% from the center and rotate it 90 degrees clockwise." 3. The assistant runs: python scripts/crop_and_rotate.py <input_path> <output_path>. 4. The result appears in ~/.open-skills/assets/outputs/. You can also run the script manually.
What are this skill's strengths and limitations?
- Fully local execution — data never leaves your Mac
- Minimal implementation depending only on PIL/Pillow; predictable behavior
- Prints dimensions at each stage for easy verification
- Supports common formats: JPEG, PNG, GIF, BMP, TIFF, etc.
- Operations are hardcoded: only 50% center-crop plus 90-degree clockwise rotation, no parameters
- Processes one image per invocation; the README's batch-processing example describes the collection's general capability, not a documented feature of this script
- Relies on OpenSkills' /mnt/user-data path conventions; outside that sandbox, paths must be adjusted by hand
- No test suite or standalone documentation independent of OpenSkills
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Image Crop and Rotate this page | 50 · Use with care | ★ 452 | 9mo ago | Apache-2.0 |
| Slack GIF Creator Toolkit | 50 · Use with care | ★ 3.1k | 8mo ago | MIT |
| TypeUI (Hermes Agent Design Skill) | 59 · Recommended | ★ 2k | 3mo ago | NOASSERTION |
| TypeUI for Grok | 59 · Recommended | ★ 2k | 3mo ago | NOASSERTION |
| TypeUI (OpenClaw Skill) | 59 · Recommended | ★ 2k | 3mo ago | NOASSERTION |
The source positions OpenSkills as a local alternative to Anthropic's official skills, which run in the cloud; this skill serves as the local image-processing option. No other competitors are named.
How did FollowSkills review this skill?
The script only reads/writes local images: no network, no credential access, no external side effects; code is fully auditable, paths are user-supplied, and errors print diagnostics. Deducted for: no user-confirmation mechanism, no rollback guidance for overwriting outputs, undeclared Pillow version inside the skill, and unverified publisher identity.
Self-consistent script: argument validation, input-file and output-directory checks, exception capture with traceback, explicit exit codes; docs match script behavior (crop math, ROTATE_270 as 90° clockwise). Deducted for: no execution or test evidence, missing edge-case handling (corrupt/non-image files, EXIF orientation via Image.open), so it sits at the static cap of 10.
Scenario is clear (center 50% crop + 90° clockwise rotate) and triggers are precise, limiting false invocation. Deducted for: fixed, non-parameterizable operations with no declared non-fit boundary (any other ratio/angle is out of scope but not stated), no Chinese support, hardcoded /mnt/user-data sandbox paths, and limited environment-fit evidence.
Readable structure (Overview / When to Use / Process / Example) and clear Apache-2.0 license. Deducted for: no skill-level versioning or changelog, no bundled requirements.txt (Pillow must be inferred from the repo root file), maintenance responsibility tied to an unverified publisher, and no FAQ or known-limitations section.
Goal is explicit and the simple script plausibly produces a directly usable rotated/cropped image. Deducted for: no execution verification of outputs, thin marginal value over a one-line PIL command, and inflexible fixed 50%/90° behavior that may force manual edits.
Source code is fully auditable primary material. Deducted for: no test suite, no committed execution evidence (CI only builds Docker images and does not cover this skill path), and no sample outputs or third-party corroboration, so even the static cap of 5 is unreachable; scored 3.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →