Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only Word formatting helper; its activation wording is overly broad, but it shows no hidden execution, persistence, credential use, or data exfiltration.
Before installing, be aware that this helper may activate on broad productivity or automation wording, so it is best used explicitly for Word/DOCX formatting, style cleanup, and find/replace tasks. It does not appear to install code or access credentials, but users should still review any generated scripts before running them on sensitive documents.
The trigger sentences are extremely broad and unnatural, but still overlap with ordinary user requests for help, which can cause the skill to activate outside its intended scope. Overbroad activation is dangerous because it can hijack unrelated conversations, suppress more appropriate skills, or steer users into generic workflows without clear consent.
The trigger phrases are generic and map to broad productivity requests, which can cause the skill to activate outside the user's clear intent. In an agent system, overly broad activation increases the chance of misrouting user requests, unexpected instruction injection from unrelated contexts, or execution of the wrong workflow on sensitive documents.
The trigger phrases are very broad and can cause the skill to activate for loosely related requests, increasing the chance of inappropriate invocation or prompt-surface expansion. In an agent environment, overbroad activation can route unrelated user content into this skill, which may lead to unintended actions, confusion, or abuse of downstream capabilities even if the README itself contains no executable code.
The skill description is broad enough to match many ordinary productivity requests, which can cause unintended activation and routing collisions with other skills. While not directly enabling code execution or data exfiltration, overbroad triggering can misroute users into workflows that are less appropriate, reducing safety and reliability of agent behavior.
The keyword list contains generic terms like 'styles' and 'automation' that appear in many unrelated conversations, making activation ambiguous. This increases the chance of accidental invocation, prompt-routing errors, and inappropriate application of this skill outside its intended Microsoft Word/DOCX formatting domain.
The example trigger sentences use everyday phrasing that is too general and likely to overlap with unrelated user requests. In agent systems, broad examples can train or bias routing toward this skill in situations where it is not the best fit, causing confusion, reduced answer quality, or unintended handling of user tasks.
The description and activation guidance are broad enough that the skill may be invoked for many general productivity or document-related requests, not just narrowly scoped Word/DOCX formatting repair tasks. Over-broad routing can expose users to irrelevant automation, increase unintended handling of sensitive documents, and reduce operator control over when the skill is applied.
The keyword list uses generic terms like 'automation', 'styles', and 'find replace' without contextual constraints, making accidental invocation likely. This can misroute unrelated requests into the skill, creating unsafe or low-quality outputs when the user's intent falls outside the skill's actual competence or expected data handling boundaries.
The default_prompt is overly broad and generic, combining common productivity terms with implicit invocation enabled. This can cause the skill to activate for routine Word or document-related requests even when the user did not clearly intend to invoke it, increasing the chance of unintended routing, prompt interference, or over-collection of context.
The invocation description and trigger block do not define a precise boundary for when the skill should run, and the sample triggers are truncated and vague. This ambiguity increases the chance of unintended routing and prompt capture, especially in a broad productivity context where many user requests partially resemble the listed keywords.
The file list presents English and Chinese instruction and guide files, but the README does not state how the user's preferred language is selected or that language choice is optional. This can create a locale-policy concern if the skill defaults users into a language without explicit opt-in.
This is a Chinese README, but key user-facing content such as the workflow description and trigger phrases is presented in English only. That creates a language-policy concern because the file implicitly forces a specific language for activation and usage without user opt-in or a documented rationale.
The file is a zh-CN skill document, but the trigger keywords and example invocations are presented only in English. This can amount to a language policy issue because it implicitly requires English phrasing for activation without stating that users may use Chinese or choose their preferred language.
No suspicious patterns detected.
