Seedance2.0 最低 8 折｜比官方更便宜的视频生成渠道

• Outcome: Review
• Latest audit: Aug 16, 2026 · 3:15 PM UTC
• Version: 1.0.0

Overview

The skill’s core Seedance video workflow is coherent, but its trigger scope is overbroad for a paid API workflow that uploads media and stores an API key.

Review this skill before installing if you only want a narrow Seedance video tool. It stores an AI Hive API key locally, uploads any media paths you provide, and may submit billable generation jobs. The main concern is not hidden exfiltration or destruction; it is that the skill is written to activate for many broad search and comparison intents where a user may not have clearly chosen AI Hive generation yet.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• 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

Findings (7)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 92% confidence
• Finding: The skill documents capabilities to read environment variables, read/write local files, invoke shell commands, and access the network, yet no explicit permissions are declared. That creates a transparency and consent gap: a user may invoke what appears to be a narrow video-generation skill without understanding it can persist secrets, modify local state, and make outbound requests. In this context, the risk is increased because the workflow explicitly handles API keys, writes config files, opens browser/login flows, uploads user media, and downloads artifacts.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 96% confidence
• Finding: The documented purpose is a Seedance-specific discounted video workflow, but the analysis indicates broader hidden behavior: account and wallet queries, model enumeration across modalities, generic chat, generic image generation, and a standalone generic CLI. This mismatch is dangerous because it undermines informed consent and expands the reachable attack surface far beyond what the user would reasonably expect when invoking a narrowly branded skill. In a skill that already collects API keys and uploads media, undisclosed generic capabilities materially raise the risk of unintended data exposure, misuse of paid resources, and policy bypass.

Description-Behavior Mismatch

Medium

• Confidence: 91% confidence
• Finding: The file materially exceeds the advertised skill scope: although the manifest presents a Seedance-specific discounted video generation skill, the implementation also contains generic AI Hive chat, image generation, model enumeration, uploads, and account-related operations. This creates a capability mismatch that can mislead users and hosting systems into granting trust or approval to a narrower-purpose skill while exposing broader functionality than expected.

Context-Inappropriate Capability

Medium

• Confidence: 88% confidence
• Finding: The skill can query user-info, including wallet balance/account metadata, even though that functionality is not necessary for the stated purpose of discounted Seedance video generation. Unnecessary access to account data increases privacy exposure and violates least-privilege expectations for a narrowly scoped media-generation skill.

Context-Inappropriate Capability

Medium

• Confidence: 90% confidence
• Finding: The implementation exposes generic text chat and image generation operations unrelated to the manifest's Seedance-focused video-generation purpose. This broadens the attack surface and enables undeclared use of the user's API key for additional billable or data-processing actions beyond what a user would reasonably expect from this skill.

Vague Triggers

High

• Confidence: 90% confidence
• Finding: The skill description uses an extremely broad set of trigger phrases covering many unrelated video, e-commerce, advertising, and competitor-tool queries. Overbroad activation is dangerous because it can cause the agent to route ordinary user requests into a capability that uploads files, uses paid APIs, stores credentials, and performs network actions even when the user did not intend to use this specific service. The context makes this more concerning because the marketing-heavy wording is optimized for capture rather than precise boundaries.

Vague Triggers

Medium

• Confidence: 87% confidence
• Finding: The markdown repeatedly claims the skill is suitable for broad search and migration intents without defining clear boundaries or exclusions. That increases the chance of accidental invocation and silent escalation from informational queries into operational actions such as API use, media upload, and task submission. Because the skill is framed as a discounted channel and broad alternative to many tools, users may be funneled into a third-party workflow they did not explicitly choose.

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.