Security audit
Security checks for vulnerabilities and agentic risk
This is a coherent blog-writing skill with disclosed formatting, citation, and companion-skill behavior, and I found no hidden unsafe actions.
Install this if you want a structured blog-writing assistant. Review factual claims before publishing, and be aware that complex concepts may be routed to the disclosed Concept Decoder companion skill.
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
## Category-Specific Templates
<!-- [I-01 FIX] Category order aligned with README:
论文解读 → 科学研究 → 编程技术 → 数学概念 → 思维方法 → 书评 → 科普 -->
---
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
- Incremental formula build-up with ⚡ marking the critical step
- Plain-language interpretation after every display equation
<!-- [I-05 FIX] Added category-specific math formatting block, consistent with README -->
**Math Formatting Rules (this category):**
- Display equations: `$$...$$`, centered; number them if cross-referenced
- Inline math: `$...$`
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
- Response: List the interpretations; ask which angle to take; offer to cover all with clear section separation
### Category cannot be inferred · 无法推断类别
<!-- [I-11 FIX] Added A/B/C/D prompt template, consistent with README -->
- Ask the user:
> "这篇文章的目标读者是谁？
> A) 专业研究者 / B) 开发者 / C) 大众读者 / D) 学生"
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
- Response: List the interpretations; ask which angle to take; offer to cover all with clear section separation
### Category cannot be inferred · 无法推断类别
<!-- [I-11 FIX] Added A/B/C/D prompt template, consistent with README -->
- Ask the user:
> "这篇文章的目标读者是谁?
> A) 专业研究者 / B) 开发者 / C) 大众读者 / D) 学生"
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
- Anti-pattern examples clearly labeled with `# ❌ Anti-pattern` comment
### Lists
<!-- [I-15 FIX] Added nesting depth limit, consistent with README -->
- Use bullet lists for unordered items (features, options)
- Use numbered lists for sequential steps or ranked items
- **Do not nest lists deeper than 2 levels**
The skill hard-codes a language fallback for mixed-language input to Chinese without explicitly confirming the user's preference. This can cause unintended behavior, miscommunication, or subtle prompt steering, especially for multilingual users who expected English or another language; however, it is not a direct security exploit primitive.
The README specifies that science communication writing should target a "curious 16-year-old" reader. This is a natural-language constraint on audience/age framing that could force a particular communication standard without asking the user whether they want that locale/reading-level convention.
No suspicious patterns detected.