Security audit
Security checks for vulnerabilities and agentic risk
This skill uploads a user-requested workspace file to an internal Hub file service and returns a temporary link, with the behavior disclosed and bounded.
Install only if you want agents to create temporary external download links for workspace files. Treat returned links as shareable outside the workspace for about 7 days, and avoid requesting links for files containing secrets or private data unless sharing is intended.
The skill invokes a shell script to upload workspace files but declares no explicit tool scope or allowed-tools boundary. That creates an authorization gap: an agent may execute shell-capable behavior without clear least-privilege constraints, increasing the chance of unintended file access or exfiltration if the skill is triggered improperly.
The description requires '对用户只说简体中文短句', which forces a specific language regardless of user preference. This is a natural-language policy issue because it does not offer any user opt-in or language selection.
The line states '对用户只说简体中文短句。', again imposing a single language with no user choice. This reinforces a locale/language restriction that is not documented as optional or region-specific.
The skill explicitly broadens activation beyond enumerated phrases ('不限于某几个固定词，下面都算'), which makes triggering ambiguous and overbroad for a capability that uploads files to an external OSS and returns a public link. Overbroad matching raises the risk of accidental activation and unintended data exfiltration from the workspace when a user merely discusses files or sharing in a non-authorizing context.
The comments and all user-facing error/output strings in this script are written in Chinese, and there is no indication that the skill is region-specific or that users can opt into another language. That creates a natural-language locale policy concern because the skill effectively forces one language for interaction.
No suspicious patterns detected.
