Security audit
Security checks for vulnerabilities and agentic risk
This image-generation skill is not malicious, but it needs Review because its stated triggers and capabilities are broader than its simple image-download purpose.
Install only if you are comfortable with a browser-driven Doubao workflow that can use your logged-in Doubao session, send prompts to that service, and save files locally. Use it only for image-generation requests, keep output paths in a dedicated folder, and avoid treating its broad AI/orchestration wording as permission for unrelated command or file operations.
The top-level description says to use the skill for broad domains such as AI model calling, intelligent dialogue, agent orchestration, and LLM applications, far beyond image generation. Overbroad invocation language can cause the skill to activate in unrelated contexts and bring along unnecessary tool permissions and risky workflows.
The description explicitly states '支持中文交互' while the file does not offer any user opt-in or alternative language choice. This can violate language or locale policy expectations when a skill implicitly constrains interaction language without documenting that it is optional or region-specific.
The document states that the free version does not support workflow automation, but elsewhere presents '自动化执行' as a core feature. Contradictory capability statements create unsafe ambiguity about whether the skill may autonomously perform broader actions, which can lead to over-permissioning or unexpected execution behavior.
The activation guidance uses vague context triggers like design creation, UI design, poster making, and brand visuals without explicit boundaries or safe-invocation criteria. This ambiguity increases the likelihood that an agent selects the skill for tasks that do not require it, exposing browser, exec, and file-copy behaviors unnecessarily.
The skill is advertised as a narrowly scoped image-generation helper, but later claims broad capabilities including file handling, API integration, command execution, and information retrieval. This scope expansion can cause an agent or reviewer to grant more trust and tool access than users expect, increasing the chance of unintended file or system actions under the guise of image generation.
The capability matrix marks '提示词增强' as unsupported, yet the skill documentation elsewhere includes '简易提示词优化' support and a full scenario where the agent gives optimization suggestions. While 'optimization' and 'enhancement' may differ slightly, the current wording presents a practical contradiction in intended functionality.
The skill instructs downloading images and copying them into local directories, but does not prominently require explicit user consent or clearly delimit safe destination paths. Even if the action seems benign, local file writes are side effects and can overwrite files, place content in sensitive locations, or normalize unsafe file operations for a broadly triggered skill.
No suspicious patterns detected.
