Skip to content
Back to skills

Model Api

ASecurity

"Inspect and extend MMYOLO registries, model components, datasets,

  • 247 stars
  • 0 votes
  • 0 copies
  • 4 views
  • Added September 8, 2026
datatestingapi

Works with

  • cli
  • api

Security analysis

A100/100

Pro scans all 5 files and shows the line behind each finding

Scanned September 8, 2026

npx -y skills add VectorSpaceLab/AREX-Skill --skill model-api --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Model Api?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Model Api
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vectorspacelab-model-api/badge)](https://www.skillsdirectory.com/skills/vectorspacelab-model-api)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: model-api
description: "Inspect and extend MMYOLO registries, model components, datasets,
  plugins, and deploy-mode modules safely."
disable-model-invocation: true
metadata:
  disco-role: operating
license: GPL 3.0
---

# MMYOLO model API and extension router

Use this sub-skill when the task is to inspect MMYOLO registries, understand public model/dataset component APIs, register custom modules, add project-local extensions, reason about plugins, or decide whether a model can be switched to deploy mode.

## Read first

1. Read [API reference](references/api-reference.md) for registry names, default-scope behavior, public component families, and installed signature facts.
2. Read [extension patterns](references/extension-patterns.md) before creating or modifying any custom model, backbone, head, loss, assigner, transform, visualizer, or project module.
3. Read [troubleshooting](references/troubleshooting.md) when registry lookup, custom imports, channel dimensions, plugins, data preprocessors, or deploy-mode conversion fail.
4. Use [inspect_mmyolo_registry.py](scripts/inspect_mmyolo_registry.py) for safe registry/module discovery. It imports/registers modules and reports registry contents; it does not build models, load datasets, train, download, or run inference.

## Route elsewhere

- Config-family selection, `_base_` inheritance edits, model zoo config choice, and `--cfg-options` belong to `config-customization`.
- Training, testing, distributed launch, resume, AMP, and evaluation CLI behavior belong to `training-evaluation`.
- End-user image/video/folder inference and visualization demos belong to `inference-visualization`.
- ONNX/TensorRT/RKNN/MMDeploy/EasyDeploy export or checkpoint conversion belongs to `deployment-conversion`.
- Dataset layout/conversion/statistics workflows belong to `data-tools`; this sub-skill covers only dataset and transform extension APIs.

## Safe operating posture

- Prefer `mmyolo.utils.register_all_modules()` before direct `MODELS.build`, `DATASETS.build`, `TRANSFORMS.build`, or `TASK_UTILS.build` use in standalone scripts.
- In reusable library modules, do not change the global MMEngine default scope at import time. Register classes with decorators and let the calling config/script handle `custom_imports` and default scope.
- Keep custom module imports explicit with `custom_imports` or a deliberate import in the caller; a class is not discoverable by registry name until its module has been imported.
- Treat `switch_to_deploy(model)` as a destructive inference/export preparation step for modules containing MMYOLO `RepVGGBlock` branches. Do not call it before continued training.

Files in this skill

  • SKILL.md2.6 KB
  • references/api-reference.md14 KB
  • references/extension-patterns.md9.9 KB
  • references/troubleshooting.md6.9 KB
  • scripts/inspect_mmyolo_registry.py10.9 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.