Security audit
Security checks for vulnerabilities and agentic risk
This skill is a competitive-research report helper with disclosed form, template, and local test assets that fit its stated purpose.
Install only if you want a Chinese/bilingual competitive-product research workflow. Review any internal materials before adding them to the intake form or generated report, and treat the optional Chrome test script as a developer-only tool.
The declared purpose is competitive research and report generation, but the file also instructs browser automation, local asset copying, DOM interaction, and shell-based regression testing. This mismatch is dangerous because users and policy systems may trust the skill under a low-risk business-analysis label while it attempts higher-risk local execution behaviors that expand attack surface.
Referenced artifact was not completely inspected
按采集表确认的格式生成：HTML 严格使用 [report-template-pro.html](references/report-template-pro.html)；Markdown 使用同一分区结构。格式锁、事实边界、文风、自检 → [research-playbook.md](references/resea
Referenced artifact was not completely inspected
按采集表确认的格式生成：HTML 严格使用 [report-template-pro.html](references/report-template-pro.html)；Markdown 使用同一分区结构。格式锁、事实边界、文风、自检 → [research-playbook.md](references/resea
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
</style>
</head>
<!--
Competitive Product Research Report Template
=============================================
Placeholder format: __placeholder_name__
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
</div>
</div>
<!-- ========== CALLOUTS ========== -->
<div class="cos">
<div class="co insight">
<span class="co-icon">🔍</span>
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
</div>
</div>
<!-- ========== SUMMARY CONCLUSIONS ========== -->
<div class="card">
<div class="card-h">📋 Summary Conclusions</div>
<ol style="padding-left:18px;font-size:13px;color:var(--s700)">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<div class="card-h">🧠 Strategic Analysis</div>
<div class="card-sub">SWOT · Porter's Five Forces · PESTLE</div>
<!-- SWOT -->
<div style="font-size:13px;font-weight:600;color:var(--s700);margin-bottom:8px">SWOT</div>
<div class="swot">
<div class="swot-cell s"><div class="swot-label">Strengths</div>__swot_s__</div>
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<a class="pill n pill-link" href="__finding_2_ev_url__" target="_blank" rel="noopener">__finding_2_ev__</a>
</div>
</div>
<!-- Risk-level finding -->
<div class="finding risk">
<div class="dim">__finding_3_dim__</div>
<strong>__finding_3_title__</strong>
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
</div>
</div>
<!-- ========== IMPLEMENTATION ROADMAP ========== -->
<div class="card">
<div class="card-h">🚀 Implementation Roadmap</div>
<table>
The skill references capabilities that imply file access, network use, and shell-driven test execution, but it does not declare any explicit tool scope or permission boundaries. That creates an authorization ambiguity where a host or reviewer may approve a seemingly harmless research skill without realizing it can access local files, open browsers, or invoke commands.
The invocation guidance says the skill can be triggered by phrases like “帮我做 XX 和 YY 竞品调研” and especially “benchmark X against Y,” which are broad formulations that may overlap with ordinary requests for comparison rather than this specific dual-track competitive-research workflow. The file provides many positive examples but no negative examples or exclusion rules for borderline comparison requests, making activation scope less specific.
This section says the listed colloquial inputs ‘都应触发本技能’ and includes broad natural phrases like asking what a product does better or requesting material before a PRD review. Without exclusion conditions, these examples may collide with general product-analysis or strategy-assistant requests rather than uniquely identifying this skill.
The page sets lang="zh-CN", and all visible labels and prompts are written in Chinese, which indicates the skill is designed for a single language/locale. Per the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the restriction is explicitly justified, which is not present here.
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
with tempfile.TemporaryDirectory(prefix="cpr-form-test-") as profile_name:
profile = pathlib.Path(profile_name)
process = subprocess.Popen(
[
str(find_chrome()),
"--headless=new",
The HTML document declares lang="en", which fixes the report output to English. Under the stated policy, forcing a specific language or locale without user choice or a documented region-specific reason is a natural-language policy concern.
The comment 一体式仪表板 introduces Chinese-language content into an otherwise English template, while the file does not explain language selection or offer locale choice. This creates inconsistent language behavior and suggests an implicit locale decision without user opt-in.
Line L240 states that if the user asks in Chinese, the report title and section names should preferentially be localized to Chinese. Under the policy criteria, forcing or defaulting to a specific language without an explicit opt-in can be a locale-policy issue when no user choice mechanism or justification is provided.
No suspicious patterns detected.