Security audit
Security checks for vulnerabilities and agentic risk
This skill is a straightforward Alibaba Cloud DocMind helper, but users should treat document URLs and cloud credentials carefully.
Install only if you intend to send documents to Alibaba Cloud DocMind. Use least-privileged Alibaba Cloud credentials, avoid printing or committing access keys, and do not use publicly accessible URLs for sensitive documents unless the link is tightly controlled and revoked after processing.
The skill includes executable code that reads cloud credentials from environment variables, but it does not declare any tool scope, permissions, or allowed-tools boundaries. In an agent setting, that omission can cause operators or orchestrators to run a credentialed skill without clear sandbox expectations, increasing the chance of unintended secret access or misuse.
The skill explicitly tells users to keep files publicly accessible for URL submission, but it does not warn that this may expose sensitive documents to anyone with the link or broader internet access depending on hosting configuration. In a document-processing context, this raises a real confidentiality risk because users may upload private contracts, IDs, or internal files for analysis.
The skill is documented throughout as a Node.js/JavaScript DocMind integration, but the Validation section iterates over scripts/*.py and runs python3 -m py_compile. This is an active contradiction in the file's own documentation because the stated validation procedure does not validate the Node.js quickstart/script content the skill presents.
The markdown instructs users to provide access key credentials via environment variables, which is a sensitive operation, but it does not include any caution about secret handling, storage, or avoiding disclosure in logs and artifacts. For a skill description that references credentials directly, a brief warning helps reduce accidental exposure.
The workflow text appears to be generic cloud-operation guidance and refers to classifying operations as read-only or mutating, then running a minimal read-only query first. For this skill, the documented behavior is narrowly to submit DocMind jobs and poll their results, so this guidance does not match the actual intent and can mislead users about the skill's operational model.
No suspicious patterns detected.
