Aliyun Qvq

Use when visual reasoning is needed with Alibaba Cloud Model Studio QVQ models, including step-by-step image reasoning, chart analysis, and visually grounded...

Security Scan

The skill mostly looks like a small helper that builds a QVQ request payload, but there are inconsistencies (validation path mismatch and no auth/usage instructions despite claiming to integrate with Alibaba Cloud) that warrant caution before installing.

Purpose & Capability

The description claims this is for Alibaba Cloud Model Studio QVQ visual reasoning, but the package only contains a tiny request-builder script and no code or instructions to call Alibaba Cloud APIs or to provide credentials. That mismatch (claimed provider integration vs. only local payload generation) is unexplained.

Instruction Scope

SKILL.md's runtime instructions are limited to compiling and running the prepare_qvq_request.py script (which only writes a JSON with a placeholder image). However the validation section has an inconsistent pass criterion: the validation command writes output/aliyun-qvq/validate.txt but the Pass criteria references output/alicloud-ai-multimodal-qwen-vqv/validate.txt (a different path). This is an incoherent test step and should be fixed.

Install Mechanism

There is no install spec and only a small Python script is included. Nothing is downloaded or written to disk beyond running the provided script, which reduces install risk.

Credentials

The skill advertises Alibaba Cloud QVQ usage but declares no required environment variables or credentials. Real calls to Alibaba Cloud would require authentication and network access; the absence of guidance about credentials or where/when requests are sent is a gap and could lead to confusion or insecure ad-hoc handling later.

Persistence & Privilege

The skill does not request persistent/always-on presence and does not modify agent/system configs. Defaults (autonomous invocation allowed) are unchanged and acceptable for this helper skill.

What to consider before installing

This package is a small request-generator for Alibaba Cloud QVQ but has a few problems you should resolve before installing: (1) fix the validation mismatch (SKILL.md expects a different validate.txt path than the command produces); (2) be aware the script only writes a JSON with a placeholder image and does not call Alibaba Cloud nor request credentials — if you intend to actually call Model Studio you'll need to add explicit, secure instructions for providing Alibaba Cloud credentials and confirm where network requests will be sent; (3) review any future code that sends the request to ensure credentials are stored securely and images/data sent to the cloud are acceptable for your privacy needs. If you only want a local request template this is low-risk; if you expect this skill to integrate with Alibaba Cloud, treat the current package as incomplete and audit any additional code that implements network/auth behavior.

Category: provider

Model Studio QVQ Visual Reasoning

Validation

mkdir -p output/aliyun-qvq
python -m py_compile skills/ai/multimodal/aliyun-qvq/scripts/prepare_qvq_request.py && echo "py_compile_ok" > output/aliyun-qvq/validate.txt

Pass criteria: command exits 0 and output/alicloud-ai-multimodal-qwen-vqv/validate.txt is generated.

Critical model names

Use one of these exact model strings:

• qvq-plus
• qvq-max

Typical use

• Mathematical reasoning from screenshots
• Diagram and chart reasoning
• Visually grounded multi-step problem solving

Quick start

python skills/ai/multimodal/aliyun-qvq/scripts/prepare_qvq_request.py \
  --output output/aliyun-qvq/request.json

Notes

• Use skills/ai/multimodal/aliyun-qwen-vl/ for standard image understanding.
• Use QVQ when the task explicitly needs stronger reasoning over visual evidence.

References

• references/sources.md