Amazon Product to Video Script

• Outcome: Pass
• Latest audit: Sep 16, 2026 · 11:17 AM UTC
• Version: 1.0.0

Overview

This skill is a straightforward Amazon listing and review analysis workflow for producing marketing video scripts, with no evidence of hidden persistence, credential use, or destructive behavior.

Before installing, expect the agent to browse Amazon product pages and embedded reviews, then optionally create local markdown outputs for the product brief and script. Do not provide Amazon credentials; the skill correctly says login-gated review pages should not be accessed.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Rogue Agent: Self-Modification, Session Persistence
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Session Persistence

Medium

• Category: Rogue Agent
• Confidence: 60% confidence

Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md

---
name: amazon-product-analysis
description: "Turn an Amazon product link into an evidence-based short-form video script — extract listing selling points, mine real buyer language from reviews, then produce a shot-by-shot script once the user picks a content direction (product-demo / narrative / direct-response). Trigger this skill when the user drops an amazon.com or amzn.to product link and mentions short video, UGC/ad video, social content, or asks 'what kind of video would work for this product' / 'write me a video script', even if they never say the word 'skill'. This version focuses on listing + review insight and does not scrape competing viral videos for teardown (see 'Known limitations' below), and does not generate the video itself (video synthesis is a future iteration — if asked to 'just produce the video' or 'analyze similar viral videos', explain the current scope instead of forcing it)."
---

# Amazon Product → Social Video Script

Behavior Manipulation

Medium

• Category: Prompt Injection
• Confidence: 75% confidence

Finding

Subtle instructions detected that may alter agent decision-making or introduce hidden biases.

Content

Scanner excerpt · SKILL.md (reported line 37)May include surrounding context.

md

### Step 2: Pull reviews, mine authentic buyer language

**Only use the review module embedded in the product page itself — don't navigate to the separate "see all reviews" page** (URL shaped like `/product-reviews/ASIN`). The embedded review section on the product page typically already has 8-10 curated reviews (mix of positive and negative), plus Amazon's own "Customers say" AI summary (which lists positive/negative mention counts by dimension, including the negative percentage per dimension) — none of this requires login, and it's usually enough to work with. **The full reviews list page can hit an Amazon login wall depending on account/network conditions** — if that happens, never suggest the user log into their Amazon account, and never have the agent attempt to log in on the user's behalf (that's an account-security boundary this skill shouldn't touch). A smaller review sample is preferable to going anywhere near the login flow to get more. If the embedded reviews + AI summary genuinely have thin negative coverage, say so plainly in the output — "negative review sample is limited; the full reviews list requiring login was not accessed" — don't pretend the coverage was thorough when it wasn't.

Use the browser tool selected in Step 1 to open the product page and pull as much review text as you can, both positive and negative — complaints in negative reviews are often the most direct entry point (e.g. "thought it would fit my big dog but it was too small" flips directly into a "made for large breeds" hook in the video).

Vague Triggers

Medium

• Category: Not specified by scanner
• Confidence: 95% confidence

Finding

The trigger phrase at line 13, "write a video script for this product," is broad enough to match many ordinary user requests that are not specifically asking to invoke this Amazon-focused skill. This can cause unintended activation, routing general script-writing requests into a workflow that expects Amazon product links and review analysis, potentially leading to incorrect handling, user confusion, or misuse of the skill outside its intended scope.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low

• Category: Not specified by scanner
• Confidence: 82% confidence

Finding

The skill's output section recommends creating a per-product working folder containing insights.md and script.md, which is a file-write behavior. The markdown does not include any explicit user warning or disclosure that the skill may create files/artifacts in the workspace.

Content

No source excerpt is available for this finding.