other
- Location
SKILL.md:266- Finding
Livestream Templates Encourage Unverified Scarcity, Social-Proof, and Health Claims
- Content
View full analysis
Vulnerability Details
File Location:
SKILL.md, lines 266, 287, 340, and 361–362
Vulnerability Type:other: Deceptive Content Generation Guidance
Risk Level: MediumEvidence
The following is a faithful English rendering of the relevant content:
markdown | **Order-closing script** | Hunger marketing, scarcity plus reassurance | Countdown plus after-sales assurance | "Only 10 orders left! Removed in 30 seconds! Customers who ordered, reply 'ordered'." | Example: "I actually used an XXX product, which guarantees that you can enjoy food without gaining weight." | **Scarcity and reassurance** | "Place your order quickly" | "Only the final 20 orders remain! Five hundred customers have tried it for you. What are you waiting for?" | Social proof plus scarcity | | Does it contain S-level prohibited terms? | Absolute or false promises | Must contain zero | | Does it contain inducement language? | Time-limit, quantity-limit, or price-increase threats | Must contain zero |Technical Analysis
The Skill is documentation rather than executable software, so this issue does not provide code execution, elevated privileges, persistence, credential access, or direct system compromise. The risk arises when an AI agent treats the supplied examples as approved, reusable livestream templates.
The templates encourage claims involving exact remaining inventory, short removal deadlines, exact customer counts, and a guaranteed weight-related outcome. The Skill does not require these claims to be validated against an authoritative inventory system, sales records, campaign configuration, or substantiated product evidence before use.
The guidance is also internally inconsistent. Lines 361–362 require generated scripts to contain no false promises and no inducement based on time limits, quantity limits, or threatened price changes, while lines 266, 287, and 340 explicitly recommend those patterns. An agent following the ...[truncated 1413 chars]
- Remediation
View remediation
Remediation Suggestions
- Remove the guaranteed weight-related outcome and replace it with language that does not claim an unsubstantiated health effect.
- Replace hard-coded scarcity, deadline, and customer-count examples with parameters such as verified inventory, verified campaign end time, and verified completed-order count.
- Require each factual sales claim to include a trusted data source and a freshness timestamp before it can be inserted into generated content.
- Prohibit the model from inventing inventory levels, sales totals, customer counts, discounts, deadlines, or product outcomes when verified data is unavailable.
- Apply the prohibited-language and inducement checks to every generated script, including examples sourced from the Skill itself.
- Configure the compliance stage to block output rather than merely warn when it detects guaranteed outcomes, fabricated scarcity, or unsupported social proof.
- Add category-specific review for health, cosmetic, food, and weight-management products.
- Add regression tests demonstrating that the templates on lines 266, 287, and 340 are rejected unless their factual elements are verified and legally permissible.
- Resolve the contradiction between the sales templates and the zero-tolerance requirements on lines 361–362, making the compliance requirements authoritative.