Security audit
Security checks for vulnerabilities and agentic risk
This skill is a coherent AI growth-strategy assistant, but it asks to retain and reuse potentially sensitive business metrics and strategy details without enough retention, redaction, or deletion boundaries.
Install only if you are comfortable sharing growth metrics, budgets, CAC/LTV, assumptions, and strategy context with the skill. Avoid entering secrets or highly sensitive commercial details unless you have agreed where they will be stored, how long they are kept, and how knowledge-base updates are approved or deleted.
The manifest description is written partly in Chinese and defines trigger phrases and exclusions in Chinese, while later sections also require Chinese metric labels and Chinese-language operational instructions. There is no statement that the user may choose their preferred language, so the skill imposes a locale/language pattern without opt-in.
The trigger lists include generic phrases such as "what should we do", "build", "design", "improve", "fix", and "audit", which overlap heavily with ordinary conversation and many unrelated business tasks. Although exclusion scenarios are provided elsewhere, these specific router rules remain broad enough that the skill could activate or route incorrectly for non-growth requests.
The skill instructs logging unrecognized inputs and updating router-related knowledge artifacts based on conversation content, creating a standing mechanism for retention and reuse of user-provided data beyond the immediate task. Without explicit minimization, consent, redaction, and storage boundaries, sensitive business information can be persistently captured and later surfaced or reused inappropriately.
The session memory schema explicitly accumulates business context, budgets, CAC/LTV data, assumptions, past outputs, and user satisfaction across turns without stated restrictions on sensitive content or retention duration. In a growth-consulting context, these fields can contain confidential commercial intelligence, making uncontrolled cross-turn persistence a meaningful data-governance and privacy risk.
Per-turn structured traces capture routing decisions, failed checks, delivered sections, satisfaction states, and knowledge deltas, creating a detailed behavioral log of the user's interactions. In this skill's business-strategy setting, such traces can reveal internal priorities, weaknesses, metrics, and decision patterns, increasing exposure if retained, reused, or accessed by others.
The optimization protocol directs writing conversation-derived decisions, failures, industry data, and strategy patterns into persistent knowledge files, which materially increases the chance that proprietary user information will be retained and later reused outside its original context. Because this skill is designed for business growth strategy, the stored content is especially likely to include commercially sensitive insights and competitive intelligence.
This markdown file contains user-facing natural-language content exclusively in Chinese, including headings and operational guidance, but does not indicate that Chinese is optional or that the skill is region-specific. That can violate language/locale policy when users are not given a choice of language or informed of the constraint.
The file's operational text and comments are written in Chinese while the routing signals and headings are partly in English, but there is no indication that the user can choose the language or that the locale restriction is intentional. This can violate language/locale policy when a skill implicitly enforces one language for maintenance or operation without documented opt-in or justification.
The router phrases are broad, generic terms like "build," "improve," and "what should we do," which commonly appear in many unrelated requests. In a routing table for an agent skill, this can cause accidental or inappropriate activation of the growth-oriented skill, leading to misrouting, over-collection of context, or unsafe handling of requests that should go to a different specialist skill.
The required appended metrics table uses Chinese column names and descriptions as a mandatory output format. Because the skill does not offer locale selection or justify a Chinese-only reporting format, this creates a natural-language policy issue for users expecting another language.
This markdown file is entirely written in English and presents role and capability guidance without any note that language is optional or localized. Under the stated policy, forcing a specific language without user opt-in can be a natural-language policy violation when no locale justification is provided.
No suspicious patterns detected.
