Alicloud Ai Image Qwen Image

PassAudited by VirusTotal on May 12, 2026.

Overview

Type: OpenClaw Skill Name: alicloud-ai-image-qwen-image Version: 1.0.3 The skill contains a Python script (scripts/generate_image.py) with a significant security vulnerability: the 'resolve_reference_image' function reads arbitrary local files from the filesystem if a path is provided in the request, without validating that the file is an image. This could be exploited via prompt injection to exfiltrate sensitive local data by passing it as a 'reference_image' to the Alibaba Cloud API. Additionally, the script manually parses sensitive credentials from '~/.alibabacloud/credentials' and uses 'urllib.request.urlopen' to download files without scheme validation, which are risky patterns in an agent-executed environment.

Findings (0)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

NoteHigh Confidence

ASI03: Identity and Privilege Abuse

What this means

The skill may use your Alibaba Cloud DashScope API key, which can consume account quota or incur costs.

Why it was flagged

The helper can read a local Alibaba Cloud credentials profile to obtain a DashScope API key. This is expected for a DashScope image-generation provider, but it is sensitive account access and is not declared in the registry metadata.

Skill content

credentials_path = Path(os.path.expanduser("~/.alibabacloud/credentials")) ... key = config.get(profile, "dashscope_api_key", fallback="").strip() ... os.environ["DASHSCOPE_API_KEY"] = key

Recommendation

Use a scoped DashScope key, verify the selected Alibaba Cloud profile, and avoid placing unrelated high-privilege credentials where this helper may read them.

NoteHigh Confidence

ASI07: Insecure Inter-Agent Communication

What this means

Prompts and reference images may leave the local machine and be processed by the external provider.

Why it was flagged

The script can read a user-specified local reference image and include it in the DashScope API request. Sending prompts and reference images to the image-generation provider is core to the skill, but users should treat those inputs as data shared with Alibaba Cloud.

Skill content

if path.exists():
        return path.read_bytes()
...
response = ImageGeneration.call(
        model=MODEL_NAME,
        messages=messages,

Recommendation

Do not submit private, regulated, or confidential images or prompts unless your Alibaba Cloud/DashScope data handling terms allow it.

NoteHigh Confidence

ASI04: Agentic Supply Chain Vulnerabilities

What this means

Different or future SDK versions could change API behavior or dependency contents.

Why it was flagged

The setup instructions install the DashScope SDK without a pinned version. This is normal for an SDK-based integration, but it leaves behavior dependent on the package version resolved at install time.

Skill content

python -m pip install dashscope

Recommendation

Install in a virtual environment and consider pinning a reviewed dashscope version for reproducible use.

NoteHigh Confidence

ASI06: Memory and Context Poisoning

What this means

Local output files may retain prompts, image URLs, or metadata after a run.

Why it was flagged

The skill intentionally persists prompts, generated image URLs, metadata, and sample responses as local evidence. This is disclosed and useful for debugging, but those saved prompts or URLs may contain sensitive project context.

Skill content

Write generated image URLs, prompts, and metadata to `output/alicloud-ai-image-qwen-image/`.
- Keep at least one sample JSON response per run.

Recommendation

Review and clean the output directory when prompts, reference images, or generated URLs are sensitive.