AI大模型专家｜Seedance 2.5 视频生成与编辑

• Outcome: Review
• Latest audit: Aug 24, 2026 · 2:30 AM UTC
• Version: 1.0.0

Overview

The skill mostly does what it claims, but it can be auto-invoked for broad video requests while sending user prompts, media, and API-key-backed jobs to AI-HIVE.

Review before installing. Use it only when you intentionally want Seedance 2.5 video work through AI-HIVE, understand that prompts and selected media files may be uploaded to that service, and are comfortable storing or passing an AI-HIVE API key. Disable implicit invocation or require explicit confirmation if your workspace handles private media or cost-sensitive API usage.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration

Findings (8)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 90% confidence
• Finding: The skill documentation describes capabilities that require environment access, local file read/write, shell execution, and network access, yet it does not declare permissions or clearly bound those capabilities. This can mislead users and reviewers about the trust level of the skill and increases the risk of unintended credential handling, local file access, or external data transmission.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 94% confidence
• Finding: The skill is presented as a narrowly scoped Seedance 2.5 video workflow, but the analysis indicates broader behavior including arbitrary model access, account/wallet retrieval, generic media upload, and local credential initialization/storage. This scope expansion is dangerous because users may invoke the skill expecting a limited video tool while it can access unrelated services, expose account metadata, and operate as a general AI-HIVE client.

Description-Behavior Mismatch

Medium

• Confidence: 88% confidence
• Finding: The file is presented as a Seedance 2.5 video-generation/editing skill, but it contains broader AI-HIVE capabilities including text chat, image generation, model enumeration, and account-related operations. This scope mismatch increases the attack surface and can violate least-privilege expectations, allowing callers or integrators to use capabilities not implied by the skill’s declared purpose.

Intent-Code Divergence

Medium

• Confidence: 79% confidence
• Finding: The module docstring explicitly describes a general-purpose AI-HIVE tool, which contradicts the skill metadata claiming a Seedance 2.5-specific workflow. This misrepresentation can mislead reviewers and users about the code’s true capability set, making it easier to hide overbroad functionality inside a narrowly branded skill.

Vague Triggers

Medium

• Confidence: 82% confidence
• Finding: Overly broad search and invocation terms increase the chance that the skill is activated for generic video or creative requests beyond its intended scope. In practice, this can route users into a workflow that performs external uploads, credential use, and model operations when they were not seeking this specific AI-HIVE-integrated skill.

Vague Triggers

Medium

• Confidence: 80% confidence
• Finding: The applicable-user section uses expansive search phrases without clear boundaries, making accidental or overly broad invocation more likely. Because this skill handles uploads, task submission, and external API interactions, ambiguous activation increases the risk of sending user content to the service in contexts where that was not intended.

Vague Triggers

Medium

• Confidence: 84% confidence
• Finding: The search coverage section contains very generic industry and product keywords, which can cause the skill to match a wide range of unrelated requests. Given the skill’s ability to interact with external services and potentially upload media, overbroad matching raises the risk of unintended data exposure or unexpected third-party actions.

Vague Triggers

Medium

• Confidence: 93% confidence
• Finding: The manifest enables implicit invocation with no visible trigger constraints, allowing the platform to auto-select this skill in contexts the user may not have explicitly intended. Because this skill can initiate external AI-HIVE video-generation/editing workflows and persist task IDs/results, unintended invocation could cause unauthorized external requests, cost-incurring actions, and leakage of user-provided media or prompts to a third-party service.

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical

• Code: suspicious.exposed_secret_literal
• Location: SKILL.md:78