Back to skill

Security audit

ACE-Step Music Generation

Security checks for vulnerabilities and agentic risk

Overview

This music-generation skill has a coherent purpose, but its installer, local API, and Docker deployment expose users to high-impact execution and network risks that need review before use.

Review this skill before installing. Do not run the curl-to-bash installer as written; download and inspect installers first, pin versions and hashes where possible, and use a dedicated virtual environment. Avoid exposing the Docker or HTTP API beyond localhost, add authentication before agent-to-agent use, and do not send generated audio, prompts, chat IDs, tokens, or webhook URLs to Telegram, Discord, or Feishu unless that external sharing is intentional.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
Findings (5)

T03 · Remote Payload Retrieval and Execution

Error
Location
capsule.json:12
Finding

Remote Installer Is Downloaded and Executed Without Integrity Verification

Content
View full analysis
Remediation
View remediation
ace-step-deploy.sh' | shasum -a 256 -c - less ace-step-deploy.sh bash ace-step-deploy.sh ``` 5. Document the exact release version and trusted signing key. 6. Run installation with ordinary user privileges and never instruct users to add `sudo` unless a narrowly scoped operation demonstrably requires it. ]]>

T09 · Insecure Skill Coding Practices

Error
Location
ace_step_agent_server.py:91
Finding

Unauthenticated Local API Allows Shell Command Injection

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
ace_step_skill.py:75
Finding

Music Generation Interfaces Execute User Input as Dynamically Generated Python Code

Content
View full analysis
Remediation
View remediation

T06 · System Persistence

Error
Location
docker-compose.yml:21
Finding

Persistent Docker Services Publish Unauthenticated Interfaces on All Host Networks

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
ace-step-deploy.sh:34
Finding

Installation Executes Unpinned Third-Party Repository and Package Content

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • 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
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (66)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
99% confidence
Finding

This is a classic tool-parameter abuse issue: untrusted API input is fed into a shell-backed subprocess, allowing an attacker to manipulate the tool invocation itself. In the context of an HTTP server, this is especially dangerous because any local process able to reach the service can potentially achieve arbitrary code execution as the server user.

Content

Scanner excerpt · ace_step_agent_server.py (reported line 149)May include surrounding context.

python
'''
        
        try:
            result = subprocess.run(
                cmd,
                shell=True,
                capture_output=True,

External Script Fetching

High
Category
Supply Chain
Confidence
99% confidence
Finding

This is a direct external-script execution pattern: curl -fsSL https://evomap.ai/assets/ace-step-deploy.sh | bash. In a deployment skill, this is especially risky because users are primed to copy-paste installation commands, making arbitrary code execution on their machine highly plausible if the host, network path, or script content is malicious or later altered.

Content

Scanner excerpt · capsule.json (reported line 12)May include surrounding context.

json
],
  "gene": "sha256:07af290b1b4ae9e0dfaaa5e7b0f9b6888f7e763f38b1e6b524ff8c21941227ae",
  "summary": "在 macOS Apple Silicon 上完整部署 ACE-Step 1.5 AI音乐生成模型,使用MLX加速,支持本地文本生成音乐",
  "content": "# ACE-Step 1.5 Mac 部署指南\n\n## 概述\n本方案在 macOS Apple Silicon (M1/M2/M3/M4) 上部署 ACE-Step 1.5,一个开源的AI音乐生成模型。使用 Apple 的 MLX 框架进行加速,无需 GPU 即可实现高效的本地音乐生成。\n\n## 硬件要求\n- **芯片**: Apple Silicon (M1/M2/M3/M4)\n- **内存**: 16GB+ (推荐 32GB)\n- **存储**: 10GB+ 可用空间\n- **系统**: macOS 13.0+\n\n## 部署步骤\n\n### 1. 一键安装脚本\n```bash\n# 下载并执行安装脚本\ncurl -fsSL https://evomap.ai/assets/ace-step-deploy.sh | bash\n```\n\n### 2. 手动安装\n\n#### 2.1 克隆仓库\n```bash\nmkdir -p ~/workspace\ncd ~/workspace\ngit clone --depth 1 https://github.com/ace-step/ACE-Step.git ace-step\n```\n\n#### 2.2 创建虚拟环境\n```bash\npython3 -m venv ~/ace-step-env\nsource ~/ace-step-env/bin/activate\n```\n\n#### 2.3 安装依赖\n```bash\npip install --upgrade pip\n\n# Apple Silicon 加速库\npip install mlx mlx-lm\n\n# 深度学习框架\npip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu\n\n# 其他依赖\npip install transformers accelerate safetensors\npip install soundfile librosa\n```\n\n#### 2.4 安装 ACE-Step\n```bash\ncd ~/workspace/ace-step\npip install -e .\n```\n\n#### 2.5 下载模型 (~10GB)\n```bash\nmkdir -p checkpoints\n\n# VAE (336MB)\ncurl -L -o checkpoints/vae/diffusion_pytorch_model.safetensors \\\n  https://huggingface.co/ACE-Step/Ace-Step1.5/resolve/main/vae/diffusion_pytorch_model.safetensors\n\n# Qwen3 Embedding (1.1GB)\ncurl -L -o checkpoints/Qwen3-Embedding-0.6B/model.safetensors \\\n  https://huggingface.co/ACE-Step/Ace-Step1.5/resolve/main/Qwen3-Embedding-0.6B/model.safetensors\n\n# LM (3.5GB)\ncurl -L -o checkpoints/acestep-5Hz-lm-1.7B/model.safetensors \\\n  https://huggingface.co/ACE-Step/Ace-Step1.5/resolve/main/acestep-5Hz-lm-1.7B/model.safetensors\n\n# DiT Turbo (4.5GB)\ncurl -L -o checkpoints/acestep-v15-turbo/model.safetensors \\\n  https://huggingface.co/ACE-Step/Ace-Step1.5/resolve/main/acestep-v15-turbo/model.safeten
...[truncated 26 chars]

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · feishu_music_sender.py (reported line 29)May include surrounding context.

python
def run_in_venv(cmd: list, cwd: str = None) -> subprocess.CompletedProcess:
    """在虚拟环境中运行命令"""
    venv_python = VENV_PATH / "bin" / "python"
    env = os.environ.copy()
    env["PATH"] = str(VENV_PATH / "bin") + ":" + env.get("PATH", "")
    
    full_cmd = [str(venv_python)] + cmd

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · generate_and_send.py (reported line 52)May include surrounding context.

python
def run_in_venv(cmd: list, cwd: str = None) -> subprocess.CompletedProcess:
    """在虚拟环境中运行命令"""
    venv_python = VENV_PATH / "bin" / "python"
    env = os.environ.copy()
    env["PATH"] = str(VENV_PATH / "bin") + ":" + env.get("PATH", "")
    
    full_cmd = [str(venv_python)] + cmd

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · gene.json (reported line 34)May include surrounding context.

json
"constraints": {
    "max_files": 10,
    "forbidden_paths": [
      ".env",
      "secrets",
      "api_keys"
    ],

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · gene_simple.json (reported line 36)May include surrounding context.

json
"constraints": {
    "max_files": 10,
    "forbidden_paths": [
      ".env",
      "secrets",
      "api_keys"
    ],

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · AGENT_USAGE.md (reported line 20)May include surrounding context.

检查状态:

bash
curl http://localhost:8765/status

生成音乐:

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · AGENT_USAGE.md (reported line 38)May include surrounding context.

md
import requests

# 生成音乐
response = requests.post('http://localhost:8765/generate', json={
    'prompt': 'Upbeat electronic music',
    'duration': 60,
    'output_path': '/tmp/my_music.wav'

External Transmission

Medium
Category
Data Exfiltration
Confidence
70% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · AGENT_USAGE.md (reported line 38)May include surrounding context.

md
import requests

# 生成音乐
response = requests.post('http://localhost:8765/generate', json={
    'prompt': 'Upbeat electronic music',
    'duration': 60,
    'output_path': '/tmp/my_music.wav'

Internal Network Request

Medium
Category
Server-Side Request Forgery
Confidence
70% confidence
Finding

Code issues a request to a loopback, link-local, or private-range host. This can reach internal services not meant to be exposed and is a common SSRF pivot.

Content

Scanner excerpt · AGENT_USAGE.md (reported line 38)May include surrounding context.

md
import requests

# 生成音乐
response = requests.post('http://localhost:8765/generate', json={
    'prompt': 'Upbeat electronic music',
    'duration': 60,
    'output_path': '/tmp/my_music.wav'

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · AGENT_USAGE.md (reported line 126)May include surrounding context.

md
python3 skills/ace-step/ace_step_agent_server.py

# 2. 测试调用 (终端 2)
curl http://localhost:8765/status
curl -X POST http://localhost:8765/generate \
  -H "Content-Type: application/json" \
  -d '{"prompt": "test", "duration": 10}'

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The guide instructs users to send generated audio and prompt text to external services but does not clearly warn that both the media and associated metadata/content will leave the local system. This creates a real privacy and data-governance risk, especially if prompts or audio contain sensitive, proprietary, or personal information.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
81% confidence
Finding

The guide includes a concrete-looking Telegram bot token format directly in the setup steps, framed as something to obtain and use for automated sending. Even though it is presented as documentation, this is natural-language guidance that normalizes handling and reuse of sensitive access credentials in a sharable file.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The Telegram Bot example uploads a local audio file to Telegram without warning that the file, chat identifier, and related metadata are sent to an external platform. Users may unintentionally disclose sensitive content because the workflow is presented as routine automation rather than an explicit outbound-sharing step.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

The URL is part of a curl command that uploads a local audio file to Telegram, making it a true external transmission path. The danger is contextual rather than malicious: the skill is a sending guide, but absent clear warnings and consent controls, users may unintentionally exfiltrate sensitive local content.

Content

Scanner excerpt · SEND_GUIDE.md (reported line 112)May include surrounding context.

  1. 发送文件:
    bash
    curl -X POST "https://api.telegram.org/bot${TELEGRAM_BOT_TOKEN}/sendAudio" \
        -F "chat_id=${TELEGRAM_CHAT_ID}" \
        -F "audio=@/path/to/music.wav" \
        -F "caption=🎵 ACE-Step Generated"
    

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

The URL is part of a curl command that uploads a local audio file to Telegram, making it a true external transmission path. The danger is contextual rather than malicious: the skill is a sending guide, but absent clear warnings and consent controls, users may unintentionally exfiltrate sensitive local content.

Content

Scanner excerpt · SEND_GUIDE.md (reported line 112)May include surrounding context.

  1. 发送文件:
    bash
    curl -X POST "https://api.telegram.org/bot${TELEGRAM_BOT_TOKEN}/sendAudio" \
        -F "chat_id=${TELEGRAM_CHAT_ID}" \
        -F "audio=@/path/to/music.wav" \
        -F "caption=🎵 ACE-Step Generated"
    

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The Discord webhook instructions send a local file to an externally reachable webhook endpoint without clearly informing the user that the upload leaves the device and may be exposed to channel members or anyone with webhook access. This omission increases the chance of accidental disclosure of local content.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The natural-language content throughout the skill file is presented in Chinese, including operational instructions and usage guidance. Under the policy, forcing a specific language without user opt-in or a documented locale-specific justification is a language/locale policy violation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The document instructs users to configure Telegram bot tokens and Discord webhook URLs and then automatically send generated content to external services, but it does not clearly warn that doing so transmits files, metadata, and sensitive service credentials outside the local environment. This can lead to unintended disclosure of generated media, chat identifiers, or mishandling of secrets if users copy these patterns into agent automation without understanding the trust boundary.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The FAQ recommends full automation via Telegram Bot or Discord Webhook with 'no human intervention,' but it omits a warning that music files or shared links may be sent externally without further review. In an agent context, this increases the chance of accidental exfiltration of sensitive generated content or sending to the wrong recipient/channel.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

User-facing comments and help/instructional strings in this skill are written only in Chinese, including operational guidance such as installation instructions. This imposes a specific language/locale without any stated opt-in or alternative, which matches the policy's language-choice concern.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The script writes executable Python code to a predictable, world-accessible path in /tmp and then executes it. An attacker on the same system could exploit symlink or race-condition behavior to overwrite arbitrary files or replace the temporary script before execution, potentially leading to code execution in the context of the user running this agent.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The surrounding comments and help text present status as a supported interface command, suggesting it returns usable structured status information. However, get_status calls check_installation (which already prints a complete JSON object) and then appends stray fields and a closing brace, producing contradictory output rather than a valid single status object.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

Nearly all user-facing messages and usage instructions in the script are written in Chinese, which implicitly fixes the interaction locale. There is no opt-in, fallback, or documented justification that this skill is intended only for Chinese-speaking users or a China-specific environment.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

This code performs a file write by creating a new script under the chosen installation path and then marks it executable. Although the script prints progress messages, it does not specifically warn the user beforehand that it will create or overwrite this file, and there is no confirmation prompt around that filesystem change.

Content

No source excerpt is available for this finding.