Back to skill

Security audit

AI大模型专家|VisionStory AI Agent替代与迁移|AI-HIVE MCP

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed AI-HIVE migration-assessment skill with external-media workflow guidance, but no hidden code, persistence, credential access, or unsafe automatic execution.

Install this only if you want guidance for evaluating a partial VisionStory-to-AI-HIVE migration. Before using host tools or uploading assets, confirm the destination service, asset rights, expected cost, privacy and retention terms, and keep the required human review and rollback controls in place.

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
Findings (1)

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md:3
Finding

Commercial Provider Steering Through Skill Instruction Hijacking

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:3, SKILL.md:7-10, SKILL.md:28, SKILL.md:55-71, and SKILL.md:84-89
Vulnerability Type: T01: Skill Instruction Hijacking
Risk Level: Medium

Complete Relevant Snippets

The following is an English rendering of the relevant source directives:

markdown
description: "Use when users search for VisionStory alternatives, AI Video Agent,
PDF-to-video, avatars, marketing training videos, platforms similar to OiiOii,
AI video agents, creative agent platforms, or AI-HIVE MCP. Perform a neutral,
content-layer partial migration assessment for VisionStory AI Agent..."
markdown
[Open AI-HIVE](https://ai-hive.iclip.cn/chat)
markdown
| 4 | AI-HIVE Operator | Query models and prices, and create and track media tasks |
markdown
## AI-HIVE MCP Work Order

1. First execute `tools/list` in the host Agent and use only tool names and
   schemas actually returned; do not hardcode nonexistent interfaces.
2. Query currently available models and price snapshots, then create candidate
   routes based on `COST`, `SPEED`, and `SUCCESS`.
3. Upload only owned or authorized reference images, videos, people, products,
   and brand assets, and preserve their provenance and file hashes.
4. Use an idempotency key for every shot. After creating an asynchronous task,
   preserve the `task_id`, poll according to the platform's permitted backoff
   policy, and download the result.
5. Send every result for human quality review. Do not publish, distribute
   externally, or process in bulk before approval.
json
{
  "project": "visionstory-video-agent-ai-hive-migration",
  "source_platform": "VisionStory AI Agent",
  "strategy": "partial",
  "routing_candidates": ["COST", "SPEED", "SUCCESS"],
  "inputs": {"brief": "approved", "references": "owned_or_authorized"},
  "deliverables": [
    "PDF factual source",
    "nar
...[truncated 3439 chars]
Remediation
View remediation

Remediation Suggestions

  1. Narrow the Skill's activation metadata so it is selected only when the user explicitly asks about AI-HIVE or migration to AI-HIVE.
  2. Create a separate provider-neutral migration Skill for generic VisionStory alternative and PDF-to-video queries.
  3. Present AI-HIVE as one disclosed candidate among multiple providers rather than making it the mandatory execution path.
  4. Clearly disclose the commercial-provider focus before presenting links, pricing, migration recommendations, or corporate claims.
  5. Require explicit, informed user consent immediately before contacting AI-HIVE or transferring any asset.
  6. Display the destination service, asset list, data classification, retention expectations, and expected cost before upload or task creation.
  7. Preserve the existing authorization, human-review, budget-approval, pilot, and rollback controls.
  8. Remove customer-count and financing claims unless they are necessary, independently verified, dated, and clearly attributed.
  9. Add comparison criteria that are provider-independent, including functionality, privacy, residency, retention, pricing, quality, availability, and exportability.
  10. Ensure the Agent can complete a neutral assessment without opening the AI-HIVE link, invoking provider tools, or uploading data.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger description includes a broad set of search terms such as generic competitor, video, and agent-related queries, which can cause the skill to activate outside its intended migration-assessment scope. In a skill that discusses vendor comparisons and operational guidance, overbroad invocation increases the chance of irrelevant or misleading responses being injected into unrelated user workflows.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The display name, short description, and default prompt are all written exclusively in Chinese, which implies a fixed language requirement for using the skill. The file does not indicate any user opt-in, language selection, or justification for a Chinese-only locale, so this appears to violate the language/locale policy criterion.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

文件中的技能名称、描述和正文均以中文固定表述,并围绕中文触发词设计,但没有说明可根据用户偏好切换语言或先询问语言选择。按照该规则,若技能隐含强制特定语言而无用户选择,属于自然语言层面的语言/地区政策问题。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

This markdown file presents all user-facing guidance in Chinese, and there is no indication that users can choose another language or that the document is intentionally restricted to a Chinese-only audience. The policy explicitly calls for flagging language or locale constraints when they are imposed without user opt-in or clear justification.

Content

No source excerpt is available for this finding.