Security audit
Security checks for vulnerabilities and agentic risk
This is a mostly static marketing playbook bundle, but two modules tell the agent it has direct access to ad and social publishing accounts without clear approval boundaries.
Install only if you want a broad marketing playbook and will keep the agent in draft/advisory mode for ads and social media. Do not let it launch campaigns, change budgets, upload customer lists, scrape platforms, schedule posts, or publish content unless a human reviews the exact action, account, audience, spend, timing, and compliance requirements first.
Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.
- `references/marketing-ideas/SKILL.md`
- `references/marketing-psychology/SKILL.md`
## Output rules
- Prefer 80/20: biggest levers first.
- Never invent metrics or keyword volumes. If missing, label assumptions.
Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.
- Cool-down period after dismiss (hours vs. days)
- Escalating urgency over time vs. consistent messaging
- Once per feature vs. consolidated prompts
- Re-show rules after major engagement
**Dismiss Behavior**
- "Maybe later" vs. "No thanks" vs. "Remind me tomorrow"
Model output is used without validation or sanitization. Unvalidated output injected into downstream contexts (SQL, shell, HTML) enables injection attacks and arbitrary code execution.
<Head>
<script
type="application/ld+json"
dangerouslySetInnerHTML={{ __html: JSON.stringify(schema) }}
/>
</Head>
{/* Page content */}
The trigger description contains broad phrases such as "test this change" and "hypothesis," which can match many ordinary product, engineering, or discussion requests that are not actually asking to invoke an A/B testing skill. This can cause inappropriate skill activation, leading the agent to steer conversations into experimentation guidance when the user intended general advice, implementation discussion, or unrelated analysis.
Tool defaults are unsafe or overly permissive (e.g. disabled TLS verification, no authentication, world-writable permissions). Unsafe defaults widen the attack surface.
**GA4 DebugView**
- Real-time event monitoring
- Enable with ?debug_mode=true
- Or via Chrome extension
**GTM Preview Mode**
The skill description is triggered by a wide range of common SEO and comparison-related phrases without defining exclusions or tighter routing criteria. That can cause the agent to invoke this skill in loosely related contexts, steering outputs toward competitor-comparison content when the user may have intended a narrower marketing, research, or writing task.
The manifest description says to use the skill when the user wants to "edit, review, or improve existing marketing copy" and includes broad phrases like "proofread," "polish this," and especially "make this better." These are common expressions that may be used outside the intended marketing-copy context, and the file does not provide exclusion conditions or negative examples to narrow activation.
The skill description contains very broad activation phrases such as 'write copy for,' 'improve this copy,' and 'headline help,' which can match many ordinary writing requests and cause the agent to invoke this skill when a more specific or safer skill would be more appropriate. Over-broad routing increases the chance of prompt/skill hijacking at the orchestration layer, misapplication of persuasive-copy behavior, and degraded separation between general writing tasks and marketing-specific tasks.
The skill description is broadly scoped and can activate on many common email-related requests, increasing the chance this skill is selected when a more specific or safer skill would be more appropriate. Overbroad routing is dangerous because it can cause the agent to apply the wrong instructions or workflow to user requests, leading to confusion, incorrect task handling, or bypass of intended skill boundaries.
The skill description contains very broad trigger phrases such as 'form optimization,' 'form fields,' and 'contact form,' which can cause the agent to invoke this skill during general discussion rather than only when the user explicitly requests CRO guidance. Over-broad activation increases the chance of inappropriate tool/skill selection, context hijacking, or irrelevant optimization advice being injected into unrelated conversations.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
## Field-by-Field Optimization
### Email Field
- Single field, no confirmation
- Inline validation
- Typo detection (did you mean gmail.com?)
- Proper mobile keyboard
The trigger list is unusually broad for a skill-routing description and can cause the skill to activate for generic marketing or product discussions that do not specifically call for this workflow. Over-broad invocation increases the chance of inappropriate tool selection, irrelevant guidance, and prompt-scope hijacking where this skill displaces a more suitable one.
The skill explicitly recommends lead capture, follow-up sequences, and sharing anonymized results, but it provides no privacy, consent, retention, or compliance guidance. That omission can lead downstream users to collect personal data or publish derived user data without adequate notice, lawful basis, minimization, or safeguards, creating privacy and regulatory risk.
The trigger description is broad enough to activate on common terms like 'launch,' 'announcement,' or 'product update,' which can cause the skill to be selected in contexts that are only loosely related. This is dangerous because unintended activation can inject irrelevant guidance, override better-matched skills, or steer user workflows inappropriately, especially in multi-skill routing systems.
The skill description includes very broad trigger phrases such as 'marketing ideas,' 'how to market,' and 'ideas to grow,' which can cause the agent to invoke this skill for many ordinary business-advice requests. Overbroad routing is a real security and reliability issue because it lets a large, instruction-heavy skill intercept unrelated prompts, increasing the chance of prompt-shadowing, reduced user intent fidelity, or unintended influence over the agent's response behavior.
The skill description contains very broad trigger conditions such as generic terms like 'psychology,' 'persuasion,' 'decision-making,' and 'consumer behavior,' which can cause the agent to activate this skill in many contexts beyond narrow marketing use. Over-broad activation increases the chance of irrelevant or manipulative guidance being injected into conversations, and this skill is especially sensitive because it focuses on influence and persuasion techniques that could shape user interactions in unintended ways.
The skill's activation text is very broad, including generic phrases like 'optimize,' 'improve,' and 'increase conversions' that can match many ordinary marketing conversations. This can cause the skill to trigger in contexts where the user did not intend CRO analysis, leading to misrouting, irrelevant guidance, or interference with more appropriate skills.
The trigger description is very broad and includes many common paid-marketing terms, which can cause the skill to activate in situations where the user did not intend to invoke an ads-specific workflow. In this skill, unintended invocation is more concerning because the content frames the agent as having direct access to advertising accounts, increasing the chance of costly or account-impacting guidance or actions being surfaced in the wrong context.
The skill states that the agent has direct access to ad platform accounts but provides no safety boundary, approval step, or warning about making spend-driving or policy-sensitive changes. In the context of paid advertising, this is dangerous because mistaken, overbroad, or manipulated execution could launch campaigns, alter budgets, or change targeting in ways that cause immediate financial loss or account compliance issues.
The manifest description says to use the skill whenever the user wants to create or optimize several broad artifacts and also includes generic terms like "upsell" and "feature gate." Without explicit exclusions beyond public pricing pages or negative examples, this could overlap with ordinary product, UX, or monetization conversations and cause unintended invocation.
The manifest description says to use the skill when the user wants to create or optimize popups, modals, overlays, slide-ins, or banners, and also when they mention generic terms like "overlay" or "banner." These terms are broad and can arise in ordinary design or UI discussions unrelated to conversion-focused popup optimization, increasing the chance of unintended invocation.
The skill recommends personalization based on visitor data and pages visited, but does not pair that advice with a clear warning to assess consent, lawful basis, minimization, and jurisdiction-specific privacy requirements. In practice, this can encourage deployment of behavior-based targeting that relies on tracking or profiling without adequate notice or controls, creating privacy and compliance risk.
The manifest says to use the skill whenever the user wants help with pricing decisions, packaging, or monetization strategy, and also on many standalone keyword mentions such as 'pricing' or 'free trial.' While domain-related, this activation rule does not define exclusions or narrower context, so ordinary discussion containing these terms could trigger the skill unintentionally.
The skill's trigger description is very broad and includes loosely bounded phrases like 'location pages,' 'comparison pages,' and 'building many pages for SEO,' which can cause the agent to invoke this skill in contexts that are only tangentially related. Over-broad activation increases the chance of inappropriate routing, causing the model to apply the wrong workflow, generate low-quality or unsafe outputs for unrelated requests, or override a more suitable specialized skill.
The skill description includes broad trigger phrases such as 'affiliate,' 'referral,' 'ambassador,' and 'word of mouth,' which can match many general marketing conversations outside the narrow intent of designing a referral or affiliate program. Overbroad activation can cause the wrong skill to engage, leading to irrelevant guidance, context capture beyond necessity, and increased attack surface through unintended routing.
No suspicious patterns detected.