Security audit
Security checks for vulnerabilities and agentic risk
This skill does what it says: uploads user-selected local files to Qiniu Cloud and returns a public URL, with no hidden persistence or unrelated behavior found.
Install only if you are comfortable giving the skill Qiniu upload credentials and publishing selected files to a public CDN URL. Do not use it for secrets, private documents, tokens, internal reports, or files you have not explicitly reviewed; ask the agent to confirm the exact path and destination before each upload.
The description emphasizes convenience and public URL generation but does not prominently warn that uploaded files become publicly accessible via a CDN. In the context of a file-upload skill, omission of this warning materially increases the risk that users expose sensitive images, HTML, documents, or locally generated artifacts without understanding the disclosure scope.
The README advertises very broad natural-language triggers such as uploading 'to the internet' or 'to Qiniu' without strong confirmation constraints. In a skill that publishes local files to a public CDN URL, broad triggers increase the chance of accidental invocation and unintended disclosure of sensitive generated or local files.
The example phrase '把这张图片上传到网上' ('upload this image to the internet') is ambiguous everyday language that can overlap with normal conversation. Because this skill performs external publication of content, ambiguous examples can encourage an agent to invoke the skill in situations where the user did not intend public distribution.
The skill accesses sensitive environment variables for cloud credentials but does not declare an explicit tool or permission scope. That weakens reviewability and least-privilege controls, making it easier for an agent or operator to overlook that the skill can use secrets and perform external uploads.
The skill is designed to return a public URL for uploaded content, but the description does not clearly warn that uploaded files become publicly accessible. Users may provide sensitive images, HTML, or local files without understanding they will be published on the internet.
The workflow explicitly supports uploading arbitrary local files and returning a public link, yet it provides no safeguards or warnings about data sensitivity. In an agent setting, this increases the risk of accidental exfiltration of local secrets, private documents, or generated files containing sensitive information to publicly reachable cloud storage.
The manifest invites broad natural-language triggers such as "put this file online," which can match ordinary requests without making the security-sensitive action explicit. In a skill that performs public cloud upload and returns shareable URLs, overly broad triggering increases the chance of unintended activation and accidental data disclosure.
The description emphasizes uploading files and returning public CDN URLs but does not clearly warn that uploaded content may become publicly accessible. This creates a meaningful risk of users or downstream agents exposing sensitive images, documents, or HTML content without informed consent.
The script uploads an arbitrary local file path to a remote cloud storage service and prints a public URL, but the execution flow itself contains no confirmation, disclosure, or restriction around what is being exfiltrated. In an agent-skill context, this increases the risk of unintended disclosure if the agent is induced to pass sensitive local files instead of only user-intended generated artifacts.
The dependency uses a caret range (^7.15.1), which allows newer compatible versions to be installed over time instead of a single fixed release. This can introduce supply-chain risk and non-reproducible builds if a later published version of the package contains a vulnerability or malicious change. In a file-upload skill that handles cloud storage operations and credentials, dependency compromise could affect uploaded content or secrets.
"name": "tianshu-qiniu-upload-skill",
"private": true,
"dependencies": {
"qiniu": "^7.15.1"
}
}
The natural-language description and console messages are fixed in Chinese, which imposes a specific language on users without offering an alternative or opt-in. This can violate language/locale policy when the skill is expected to be generally usable across locales.
No suspicious patterns detected.