Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed Chinese tender-document risk checklist assistant with no bundled code, persistence, credential access, or hidden execution behavior.
Before installing, users should know this skill is intended for Chinese tender/procurement document review and may use configured IMA knowledge bases for legal/case references. Treat its output as a checklist for human review, not legal advice or a final bid/no-bid decision, and avoid relying on it for pricing strategy or document drafting.
The entire skill description is presented only in Chinese and does not offer any language or locale choice for users. Under the stated policy, forcing a specific language without opt-in can be a natural-language policy violation unless the locale restriction is explicitly documented and justified.
The trigger list contains broad, natural-language phrases such as '这个标能不能投' and '资格审查能不能过' that are likely to appear in ordinary bidding-related conversations, making accidental invocation plausible. In this skill context, unintended activation can expose uploaded tender documents to the wrong workflow, produce overconfident compliance-style guidance, and crowd out a more appropriate skill or general assistant response.
该技能的显示名称及全文说明均以中文撰写，且未见任何允许用户选择输出语言或根据用户偏好切换语言的说明。按规则，若技能在自然语言层面实际上强制特定语言而未提供用户 opt-in/choice，可构成语言或 locale 政策问题。
L006 的触发词中包含“风险雷达”“标书体检”“这个标能不能投”“资格审查能不能过”等较口语化、覆盖面较宽的表述，其中部分短语可出现在一般咨询场景，而不一定明确指向本技能定义的“招标文件专项风险筛查”。虽然同句补充了“当用户明确要求对招标文件做专项风险筛查时触发”，但缺少更清晰的正反例或更严格的触发约束，仍存在误调用风险。
The manifest positions the skill as a specialized extractor that scans tender documents and outputs a structured checklist, with explicit scope limits around focused risk screening. However, Step 2.5 adds a separate capability: querying multiple external knowledge bases for laws, policy interpretations, practice guides, and real cases, which goes beyond pure document-side extraction and into external legal augmentation.
The manifest excludes '报价策略与评分拆解建议', suggesting the skill should avoid pricing-related analysis. Yet the few-shot guidance explicitly analyzes whether a bid falls below a cost warning threshold and how to prepare justification materials for low-price review, which is a pricing-sensitive capability adjacent to bid strategy rather than pure rejection-clause extraction.
The manifest defines multiple natural-language trigger phrases that are broad and conversational, such as '这个标能不能投' and '资格审查能不能过', which could match ordinary procurement-related user requests beyond the intended narrow workflow. Overbroad activation can cause the skill to be invoked unexpectedly, leading to scope confusion, unintended processing of sensitive bidding documents, or interference with other more appropriate skills.
For a skill whose stated purpose is extracting rejection clauses and hidden risks from a supplied tender document, external retrieval of laws, policy explainers, and a large corpus of real rejection cases is not an obvious minimum requirement. This adds a research and legal-reference capability that is not clearly justified by the concise manifest purpose statement.
No suspicious patterns detected.
