Stable Diffusion Sd3
v1.0.2Stable Diffusion 3 and SD3.5 Large on Apple Silicon — generate Stable Diffusion images locally with DiffusionKit's MLX-native backend. SD3 Medium for fast St...
⭐ 2· 115·1 current·1 all-time
byTwin Geeks@twinsgeeks
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (local Stable Diffusion on Apple Silicon, fleet routing) align with the instructions: examples use a local router on http://localhost:11435, recommend installing ollama-herd, diffusionkit, and mflux. Declared required binaries (curl/wget, optional python3/pip) match the documented commands. The metadata's configPaths (~/.fleet-manager/...) are consistent with a fleet router but are not surprising for this purpose.
Instruction Scope
SKILL.md stays within purpose: it instructs installing a fleet router (herd/herd-node), installing backends (DiffusionKit, mflux), and calling local HTTP endpoints for image generation/monitoring. It does not ask to read unrelated user data or external secrets. Note: the instructions require downloading model weights (HuggingFace) and running install/patch scripts—these are expected for model usage but involve substantial network I/O and running third‑party code.
Install Mechanism
The skill is instruction-only (no install spec). However, the guide tells users to run pip install ollama-herd and uv tool install diffusionkit, which will fetch and execute third‑party packages/binaries at install time. This is normal for such tooling but increases the attack surface compared to a purely local-only script; users should verify the provenance of those packages and scripts.
Credentials
The skill requests no environment variables or credentials. All runtime interactions are local (localhost) or involve downloading model weights from known model hosts (HuggingFace) as part of normal operation. There are no unrelated secret requests.
Persistence & Privilege
always:false and no special privileges requested. The skill does not instruct modifying other skills or system-wide configurations beyond installing tools for the router and node components; autonomous invocation is allowed but that is the platform default.
Assessment
This skill appears internally consistent with its goal of running Stable Diffusion locally, but before installing or running anything: (1) review the PyPI package 'ollama-herd' source and any 'uv tool' provider to ensure you trust them; (2) inspect any provided patch scripts (e.g., patch-diffusionkit-macos26.sh) before executing; (3) expect large downloads (2–8GB model weights) and significant RAM usage—run on an isolated or well-backed-up machine if concerned; (4) the router opens a local HTTP port (11435) — confirm it is bound only to localhost or properly firewalled if you do not want other LAN hosts to access it; (5) if you use private HuggingFace assets, verify whether authentication is needed and handle tokens separately; and (6) consider running installs in a virtualenv or dedicated environment to limit accidental system-wide changes.Like a lobster shell, security has layers — review code before you run it.
apple-siliconvk97dt3c4757bxfy1hdnjpbz5798406gqdiffusionkitvk97dt3c4757bxfy1hdnjpbz5798406gqfluxvk97dt3c4757bxfy1hdnjpbz5798406gqimage-generationvk97dt3c4757bxfy1hdnjpbz5798406gqlatestvk97bh8xr1yq5wvfqpjycb168b1845s85local-imagevk97dt3c4757bxfy1hdnjpbz5798406gqmac-minivk977t2fk7rxsz9mbrj9tkme1vh83z9v8mac-studiovk97dt3c4757bxfy1hdnjpbz5798406gqmfluxvk97dt3c4757bxfy1hdnjpbz5798406gqmlxvk97dt3c4757bxfy1hdnjpbz5798406gqno-cloudvk97dt3c4757bxfy1hdnjpbz5798406gqsd3vk97dt3c4757bxfy1hdnjpbz5798406gqsd3.5vk97dt3c4757bxfy1hdnjpbz5798406gqstable-diffusionvk97dt3c4757bxfy1hdnjpbz5798406gqtext-to-imagevk97dt3c4757bxfy1hdnjpbz5798406gq
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
art Clawdis
OSmacOS
Any bincurl, wget