Security audit

AdsTurbo 视频精修 · AdsTurbo Video Cleanup

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent AdsTurbo video-editing wrapper, but users should be careful because it sends media to a third-party service and can remove attribution-like elements.

Install only if you are comfortable sending selected media and job metadata to AdsTurbo under your API key. Do not upload private or confidential videos unless you understand the service's storage and sharing behavior, and use watermark/logo/subtitle removal only on media you own or are authorized to modify.

A.I.G

Vulnerability Patterns

Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (9)

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 92% confidence
Finding: The skill’s activation guidance is extremely broad, matching generic phrases like 'improve quality,' 'upscale,' and 'add subtitles' that commonly appear in ordinary conversation. This can cause the agent to invoke the skill when the user did not intend to use an external paid API or to perform sensitive media manipulation such as watermark or subtitle removal, increasing the risk of unintended actions and policy-bypassing transformations.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 90% confidence
Finding: The Chinese trigger examples are similarly overbroad, using common phrases such as '去字幕', '提升画质', and '加字幕' without enough qualification. In multilingual settings this increases accidental invocation risk and may lead to unintended external API use or unauthorized media edits, including removal of logos, watermarks, or burned-in subtitles.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 96% confidence
Finding: The skill instructs users to upload local assets so they become publicly accessible URLs, but it does not warn that this may expose sensitive photos, audio, or videos to anyone with the link or to public indexing depending on the storage backend. In an agent context, this increases the chance that private user-provided files are published externally without informed consent or data-classification checks.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 88% confidence
Finding: The documentation explicitly promotes removal of watermarks and burned-in subtitles without any warning about provenance, copyright, licensing, or consent implications. That lowers friction for content laundering or unauthorized reuse, especially because watermark removal is presented as a normal enhancement workflow rather than a restricted or reviewed action.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 82% confidence
Finding: The upload method reads a local file and sends its contents via multipart HTTP POST to the AdsTurbo service. Although the function name implies upload behavior, this code path does not provide a confirmation prompt or explicit user-facing disclosure at the point where local data is transmitted.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 93% confidence
Finding: This code uploads a user-supplied local file path to remote API endpoints via client.upload, which is a network/data-transmission action covered by the warning requirement for code files. Although the module docstring explains that it turns a local file into a public URL, there is no visible confirmation prompt, user-facing log/print, or explicit disclosure near the upload actions that the local file contents will be sent to a remote service and exposed via a public URL.
Content: No source excerpt is available for this finding.

Intent-Code Divergence

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The module docstring asserts that 'Nothing in frame changes meaning,' yet the implemented features explicitly remove watermarks, logos, props, and burnt-in subtitles. Those operations can materially alter provenance, attribution, and the semantic interpretation of the video, creating a risk of deceptive or misleading output and unsafe operator assumptions.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 92% confidence
Finding: The script forwards video URLs, workspace IDs, callback IDs, and related processing metadata to remote API endpoints, but this file provides no user-facing disclosure or consent warning. If users supply sensitive or private media, they may unknowingly transmit it to an external service, creating privacy, compliance, and data-handling risk.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

Category: Not specified by scanner
Confidence: 82% confidence
Finding: The file switches from English to Chinese and duplicates the full instructions in both languages, but it does not state whether the skill will communicate in one language, both, or based on user preference. This can create a locale-policy issue if the skill assumes or forces a language without explicit user choice.
Content: No source excerpt is available for this finding.