Security audit
Security checks for vulnerabilities and agentic risk
This is a markdown-only SEO page-planning skill with disclosed guardrails, though users should understand it may save generated business/SEO work to project memory and can guide public publishing or URL submission when explicitly used.
Install this if you want an SEO page-system planner and are comfortable with generated deliverables, competitor facts, local business details, and publish blockers being saved into project memory when used. Review outputs before publishing, especially competitor claims, third-party platform posts, GBP/NAP details, and any live URL submission using API keys.
The display name, summary, and description prominently include Chinese-only phrasing alongside English trigger text, and the activation text explicitly says the skill also fires on Chinese phrases. There is no statement offering a language choice or clarifying that language follows the user's preference, which can violate language/locale policy requirements.
The trigger description includes broad phrases like "how do we compare to X" and "battle card page," which can match common business or research requests without clearly constraining that this mode is only for publishing comparison pages. In an agentic system, over-broad routing can send users into a page-construction workflow that encourages creation of SEO-targeted competitor pages, including reputation-sensitive comparison content, when a narrower research, analysis, or sales-enablement skill would be safer and more appropriate.
The activation text lists general intents such as wanting to "optimize for local search" or "improve a Google Business Profile," which are broad enough to overlap with many ordinary SEO requests. The file does not provide exclusion conditions or negative examples to distinguish when this mode should fire versus other SEO modes.
The skill repeatedly states that it writes deliverables and summaries to persistent memory paths such as memory/content/, memory/hot-cache.md, and memory/open-loops.md, but it does not clearly surface a user-facing warning that generated content and project state may be retained. In an agent environment, silent persistence can expose sensitive business data, URLs, competitive intelligence, or local business details beyond the immediate task context.
No suspicious patterns detected.
