用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/Ntizar/MasterMind --skill depth-anything-3命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
| name | depth-anything-3 |
| description | Usa a estimar profundidad con Depth Anything V3. |
| version | 2.0.0 |
| tags | ["profundidad","depth-anything","vision","monocular","transformers","v3"] |
| related_skills | ["depth-anything-3","threejs-webgpu-relighting","mlx-vlm-inference"] |
⚠️ Corrección 2026-09-05 (auditoría): la API real es
depth_anything_3.api.DepthAnything3+from_pretrained+model.inference(); nodepth_anything_v3.dpt.DepthAnythingV3/predict(). El checkpoint HF es el del repo (ByteDance-Seed).
Repo: https://github.com/ByteDance-Seed/Depth-Anything-3 (Python, ~6.3K⭐).
from depth_anything_3.api import DepthAnything3
model = DepthAnything3.from_pretrained("ByteDance-Seed/Depth-Anything-3", device=...)
depth = model.inference(image_path) # método inference (no predict)
depth_anything_3.api.DepthAnything3; no depth_anything_v3.dpt.DepthAnythingV3.model.inference(...), no predict().DepthAnything3.from_pretrained(...) → model.inference(img) y comprobar el mapa de profundidad.Restore Hermes from a backup repo or migrate machines.
Usa al analizar ADN crudo: ascendencia, rasgos, sexo.
Usar al operar Mastermind en Windows: repo, ChromaDB, crons, gateway Telegram (colapso por token revocado y su restauración), backups y monitorización.