Security audit
Security checks for vulnerabilities and agentic risk
This skill is a coherent research-figure workflow guide with image-generation steps, and the main issue is a rigid Chinese-language response format rather than hidden or unsafe behavior.
Install this if you want a structured research-paper inspiration-figure workflow and are comfortable with image-generation service use. Expect Chinese section headings and suggested prompts unless you manually adapt the skill, and avoid providing proprietary paper content or reference images unless you are allowed to send them to the chosen image-generation provider.
The skill hard-codes Chinese section headers and required prompt text in every text response, without any language negotiation or fallback. This can override the user's preferred language, reduce usability, and in some integrations cause downstream prompt-routing or review errors if other components expect the conversation language to remain consistent.
The skill hard-codes required Chinese section headers in every text-only reply without any user language negotiation. This can override the user's preferred language, reduce transparency, and cause users to misunderstand workflow state or safety-relevant instructions, especially in multilingual or accessibility-sensitive contexts.
The example hardcodes Chinese-only workflow text and next-step prompts without offering any language choice, which can override or conflict with a user's preferred language and reduce transparency about how the skill behaves. In an agent setting, forced language constraints can cause misunderstanding, incorrect approvals, or inaccessible outputs, especially when users cannot reliably review generated prompts or state transitions.
The example content is written entirely in Chinese and includes prescribed follow-up prompts in Chinese without offering any language selection or fallback. This can override or constrain user interaction language unexpectedly, causing usability failures, misinterpretation of instructions, and reduced accessibility for users who do not read Chinese.
The file uses prescriptive natural-language instructions like 'The actual P5 reply must contain only generated image artifacts,' 'no prose,' 'no caption,' and 'The next TEXT_ONLY reply... must record... and review the candidates.' These constraints force a specific communication format and implicitly exclude user language/locale choice without documenting any opt-in or justified regional/compliance need.
The example startup turn is entirely in Chinese and directs the interaction flow without offering any language preference check or opt-in. This can mislead or exclude users who operate in another language, increasing the risk of misunderstanding workflow state, prompts, or generated figure requirements in a research context where precision matters.
The manifest hard-codes broad Chinese activation prompts such as a recommended prefix and fallback next-step prompts without scoping when they should be used. In a generic Chinese conversation, these strings can cause unintended skill routing or over-activation, which may steer the assistant into this workflow even when the user did not clearly request it.
The skill prescribes Chinese-language prompt prefixes and next-step text without offering a language-neutral alternative or documenting a required Chinese-only deployment scope. This can coerce interaction patterns, create confusion for non-Chinese users, and increase the chance of accidental or opaque activation through language-specific routing behavior.
The workflow hard-requires Chinese section headings and a Chinese next-prompt in every text reply, with no language negotiation or fallback. This can override the user's requested language, create misleading or inaccessible output, and act as a prompt-level quality/safety issue by reducing transparency and usability for users who cannot read Chinese.
The template explicitly mandates a Chinese section header and footer content in every TEXT_ONLY reply, which can override the user's language preference and produce unexpected locale-specific output. In an agent skill, hard-coded language constraints can degrade usability, confuse users, and create prompt-channel mismatch that weakens trust and compliance with user instructions, though it is not a direct code-execution or data-exfiltration issue.
Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.
permit persons to whom the Software is furnished to do so.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
The README content is entirely in Chinese and does not indicate that other languages are supported or that Chinese is required for a specific regional or compliance reason. Under the policy rule, forcing a specific language without user opt-in is a natural-language policy concern.
This markdown file includes operational instructions such as "当前执行计划" and later user prompt templates in Chinese, which effectively steer interaction in a specific language. The file does not indicate that language is optional or user-selectable, so it may violate the policy against forcing a specific language without opt-in.
The section "下一步你可以这样问" provides only Chinese prompt examples for invoking the skill. Because no alternative language option or opt-in is provided, the file appears to prescribe a specific language for use rather than letting the user choose.