Security audit
Security checks for vulnerabilities and agentic risk
The skill appears to be a lightweight planning/workflow skill, but its name says USA business migration while its content is actually about Qwen Image 3 product and SEO advice, with broad implicit triggers.
Install only if you specifically want Qwen Image 3 product/SEO workflow help, not USA business migration planning. The publisher should rename and rescope the skill and narrow implicit triggers before normal use.
The skill’s declared identity ('USA Business Migration Planner') does not match its documented function, which is centered on Qwen Image 3 product/SEO advice. This kind of semantic mismatch can cause incorrect routing, user deception, or policy bypass if systems or users rely on the name/manifest to decide when the skill should be invoked.
The trigger phrases are broad and loosely scoped, combining generic verbs ('help me', 'I need a practical workflow') with overloaded keywords. This increases the chance of accidental or inappropriate activation, which is especially risky here because the skill is already misrepresented and may intercept unrelated user requests.
The skill's identity is materially inconsistent: its name and heading imply a USA business migration planner, while the manifest and body target product/SEO advice for a Qwen Image 3 tool site. This mismatch can cause incorrect activation, mislead users and orchestrators, and route unrelated requests into an unintended workflow, which is a genuine security and safety boundary problem for agent skills.
The activation guidance in the description is ambiguous and overbroad, telling the system to use the skill for generic needs like practical workflows, artifacts, checklists, analysis, or implementation support. In context, this makes the skill eligible for a wide range of unrelated tasks, amplifying the risk of inappropriate routing and unexpected behavior.
The trigger keywords are extremely broad and generic, including terms like 'create', 'image', and 'seo', which are common in ordinary user requests. This increases the chance of accidental or excessive skill invocation, causing cross-domain prompt injection-like behavior where an unrelated skill captures requests it should never handle.
The skill’s declared identity and its actual behavior are materially inconsistent: it presents as a USA business migration planner while the body is about Qwen Image 3 product/SEO guidance. This can cause incorrect skill routing, mislead users and downstream agents, and hide the true operational scope of the skill, which is dangerous in systems that rely on metadata for invocation and trust decisions.
The inline documentation explicitly labels the skill as a USA Business Migration Planner while simultaneously documenting an unrelated Qwen Image 3 SEO/product task. This contradiction increases the chance of misclassification and unintended activation, and in an agent environment it can conceal the real function of the skill from reviewers or policy controls.
The trigger keywords are overly broad terms such as 'create', 'image', and 'seo', which can match many unrelated user requests. In an agent system this may invoke the wrong skill unexpectedly, causing context hijacking, user confusion, or inappropriate outputs outside the intended domain.
The default prompt uses a very broad invocation phrase tied to a vague, overloaded description, which can cause the skill to be auto-invoked for loosely related requests. In this skill, the mismatch between the display name ('USA Business Migration Planner') and the embedded task text about a Qwen Image 3 tool site increases the risk of incorrect routing, user confusion, and unintended exposure of the skill in unrelated contexts.
The skill’s declared identity and its actual documented behavior are materially inconsistent: the metadata says it is a USA business migration planner, while the requirement plan is about Qwen Image 3 product and SEO advice. This can cause misrouting, conceal the true purpose of the skill from reviewers and callers, and undermine policy or permission boundaries that depend on accurate labeling.
The README presents core requirement text in Chinese inside an otherwise English document, but it does not explain any language-selection behavior or offer user opt-in for language/locale handling. This can indicate an implicit language policy that may not align with user expectations.
The README title and stated skill identity claim this is a 'USA Business Migration Planner', but the actual documented demand and workflow are for Qwen Image 3 product/SEO advice. This mismatch can mislead users and routing systems about the skill's true purpose, increasing the chance of inappropriate invocation, unsafe delegation, or trust based on a false domain label.
The trigger phrases are broad natural-language templates that can match common user requests, causing the skill to activate outside its intended scope. In an agent ecosystem, overbroad triggers can hijack unrelated conversations, inject irrelevant workflows, and increase the risk of unintended tool use or misrouting.
The example triggers use generic formulations like 'Help me' and 'I need a practical workflow,' which provide weak constraints and encourage broad matching. While less severe than the keyword list itself, these examples still train selection logic toward over-activation and can contribute to skill misuse in adjacent contexts.
The manifest description and body are primarily in Chinese, while key trigger examples and embedded demand text are in English, and there is no instruction that the assistant should adapt to the user's preferred language. This can create a locale/language policy issue because the skill does not explicitly offer language choice or justify the mixed-language requirement.
The default prompt includes Chinese text alongside English, which can impose a language/locale assumption without any opt-in or documented reason. The file does not indicate that users can choose their preferred language or that the mixed-language behavior is required for a region-specific purpose.
The trigger sentence uses a very broad natural-language pattern ('Help me ...') that can match ordinary user requests far outside the intended scope. Overbroad activation increases the chance the wrong skill is invoked, which can produce irrelevant guidance, bypass user intent, or create an opening for stealthy prompt-squatting when combined with misleading metadata.
This trigger relies on a vague workflow-request pattern rather than a precise domain-specific request, making unintended activation likely. In the context of already-misaligned skill metadata, this ambiguity is more dangerous because it broadens the surface for accidental or deceptive invocation of a skill whose true function is unclear.
This user-facing documentation contains a mixed-language requirement statement, which may force a particular language experience on users who selected the zh-CN README expecting consistent localization. The file does not state that mixed-language content is intentional or provide an opt-in or alternative for language preference.
