Back to skill

Security audit

Meeting Transcriber

Security checks for vulnerabilities and agentic risk

Overview

This meeting transcription skill is mostly coherent, but it runs its core transcription code from an unaudited fixed local path and makes persistent setup changes, so it needs review before installation.

Install only if you trust and have reviewed the external D:/dev/python/voiceFunAsr/vocie_mic_fixed_gbk.py script and the local Conda environment. Treat meeting audio and transcripts as sensitive, get participant consent before recording, and avoid running the setup script until you understand the files it creates or modifies.

Vulnerability Patterns
  • Tool Hijacking and SpoofingModifies or replaces tools so legitimate-looking calls execute attacker logic
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T07 · Tool Hijacking and Spoofing

Error
Location
transcriber.py:94
Finding

Execution of an Unverified External Transcription Script

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
transcriber.py:473
Finding

Unpinned Third-Party Dependency Installation Instructions

Content
View full analysis
=3.8 - pip # Packages installed through pip - pyaudio - webrtcvad - sounddevice - funasr>=1.0.27 - torch>=1.13.1 - torchaudio>=0.13.1 # Optional enhanced functionality - openai-whisper - pandas - scikit-learn """ ``` The project documentation also recommends the following installation command: ```text pip install funasr torch pyaudio webrtcvad ``` ### Technical Analysis The dependency installation guidance allows pip to resolve whichever package releases are available from the configured index at installation time. No lock file, exact version set, package hashes, or explicit trusted package index is supplied. Minimum-version constraints such as `funasr>=1.0.27` do not provide reproducibility or integrity because later releases remain eligible. Packages with no version constraints are even less controlled. If an upstream account, package release, package index, dependency, or local pip index configuration is compromised, malicious installation or import-time code can enter the environment. The current `setup_environment()` implementation prints these commands rather than automatically running them. Exploitation therefore requires a user or administrator to follow the documented setup procedure, but the instructions are part of the intended operational workflow. ### Attack Path 1. A user follows the S ...[truncated 1266 chars]
Remediation
View remediation
=`; use exact versions that have been tested and audited. 7. Regularly scan the locked dependency graph for known vulnerabilities and review updates before changing pins. 8. Run installation and transcription under a dedicated, non-administrative account with access limited to the required microphone, model cache, and transcript directory. ]]>
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (36)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
99% confidence
Finding

Using shell=True for date /T && time /T is unnecessary and risky because it delegates execution to a shell interpreter. In a skill or automation context, unnecessary shell use is especially dangerous since such scripts may be run with user trust and broad filesystem access.

Content

Scanner excerpt · conda_setup.py (reported line 180)May include surrounding context.

python
"python_path": python_path,
            "env_path": env_info.get('audioProject_path'),
            "conda_version": env_info.get('conda_version'),
            "detected_at": subprocess.run(
                'date /T && time /T',
                shell=True,
                capture_output=True,

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 · openclaw_wrapper.py (reported line 107)May include surrounding context.

python
bufsize=1,
                universal_newlines=True,
                cwd=r"D:\dev\python\voiceFunAsr",
                env={**os.environ, "PYTHONUNBUFFERED": "1"}
            )
            
            print(f"转录进程已启动,PID: {process.pid}")

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill advertises real-time microphone transcription and automatic file saving, but does not clearly warn users that spoken meeting content will be captured from the microphone and persisted to disk. This creates a privacy and consent risk, especially in meetings containing sensitive, personal, or confidential information, because users or bystanders may not realize recording/storage is occurring.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The natural-language description states the skill supports Chinese speech recognition, and the keywords reinforce a Chinese-only locale focus, but there is no indication that users can choose another language or that the restriction is a justified region-specific requirement. This can violate language/locale policy when a skill forces a specific language without opt-in.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The command triggers are short, natural-language phrases that can overlap with ordinary user requests, increasing the chance of accidental invocation or prompt confusion. In an agent setting, this can cause unintended execution of transcription-related actions, including opening files or starting/stopping processes, without clear user intent.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · check_env.py (reported line 23)May include surrounding context.

python
# 获取版本
        try:
            result = subprocess.run(
                [str(python_exe), "--version"],
                capture_output=True,
                text=True,

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · check_env.py (reported line 101)May include surrounding context.

python
all_ok = True
    for dep, description in deps:
        try:
            result = subprocess.run(
                [python_exe, "-c", f"import {dep}; print('OK')"],
                capture_output=True,
                text=True,

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · transcriber.py (reported line 367)May include surrounding context.

python
all_ok = True
    for dep, description in deps:
        try:
            result = subprocess.run(
                [python_exe, "-c", f"import {dep}; print('OK')"],
                capture_output=True,
                text=True,

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · check_env.py (reported line 127)May include surrounding context.

python
print("\n[4] 检查音频设备...")
    try:
        result = subprocess.run(
            [python_exe, "-c", """
import pyaudio
p = pyaudio.PyAudio()

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · transcriber.py (reported line 395)May include surrounding context.

python
print("\n[4] 检查音频设备...")
    try:
        result = subprocess.run(
            [python_exe, "-c", """
import pyaudio
p = pyaudio.PyAudio()

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The user-facing instruction tells the user to say the fixed Chinese phrase '开始会议转录' in OpenClaw, which imposes a specific language for activation. The file does not provide an opt-in, alternative language, or explicit justification for this locale constraint.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · conda_setup.py (reported line 30)May include surrounding context.

python
env_info['detected_by'] = 'CONDA_PREFIX'
        
        # 方法2: 运行 conda info 命令
        result = subprocess.run(
            ['conda', 'info', '--json'],
            capture_output=True,
            text=True,

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · conda_setup.py (reported line 123)May include surrounding context.

python
# 测试 Python 版本
        try:
            result = subprocess.run(
                [python_path, '--version'],
                capture_output=True,
                text=True,

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · conda_setup.py (reported line 143)May include surrounding context.

python
dependencies = ['pyaudio', 'webrtcvad', 'funasr', 'torch']
        for dep in dependencies:
            try:
                result = subprocess.run(
                    [python_path, '-c', f'import {dep}; print(f"{dep} version: OK")'],
                    capture_output=True,
                    text=True,

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This section writes configuration under the user's home directory and later updates skill assets without clearly warning the user beforehand about all filesystem changes. Silent writes to persistent config and project files are risky in a skill context because users may assume the script is only performing environment checks.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
98% confidence
Finding

This code uses subprocess.run(..., shell=True) for a simple timestamp command, which invokes the system shell unnecessarily. Even though the string is currently constant, shell execution increases attack surface and can be abused via shell resolution behavior or future modifications that introduce variable input.

Content

Scanner excerpt · conda_setup.py (reported line 180)May include surrounding context.

python
"python_path": python_path,
            "env_path": env_info.get('audioProject_path'),
            "conda_version": env_info.get('conda_version'),
            "detected_at": subprocess.run(
                'date /T && time /T',
                shell=True,
                capture_output=True,

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The function claims to update skill metadata, but it also rewrites meeting_minutes.py and creates a new launcher batch file. Modifying executable skill files broadens behavior beyond what a user may expect from setup and can silently introduce persistence or execution changes in the skill.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · conda_setup.py (reported line 369)May include surrounding context.

python
- scikit-learn

# 环境配置说明
# 1. 创建环境: conda create -n audioProject python=3.8
# 2. 激活环境: conda activate audioProject
# 3. 安装依赖: pip install -r requirements.txt
# 4. 验证安装: python -c "import funasr; print('FunASR OK')"

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The instruction tells the user to invoke the skill using a Chinese phrase, with no indication that other languages are supported or that the language choice is optional. This is a natural-language locale constraint presented without opt-in, which can violate language-choice policy expectations.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This Python file contains all docstrings and interactive prompts in Chinese, including the title and menu shown to the user. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is clearly documented and justified, which is not present here.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · openclaw_wrapper.py (reported line 42)May include surrounding context.

python
print()
            
            cmd = ["cmd", "/c", "start", "cmd", "/k", str(batch_file)]
            subprocess.Popen(cmd)
            
            print("转录系统已在新窗口中启动")
            print("请在新窗口中查看实时转录输出")

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · openclaw_wrapper.py (reported line 74)May include surrounding context.

python
print()
            
            cmd = ["cmd", "/c", "start", "cmd", "/k", str(batch_file)]
            subprocess.Popen(cmd)
            
            print("转录系统已在新窗口中启动")
            print("请在新窗口中查看实时转录输出")

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · transcriber.py (reported line 20)May include surrounding context.

python
"""运行命令并处理输出"""
    try:
        if capture_output:
            result = subprocess.run(
                cmd,
                capture_output=True,
                text=True,

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · openclaw_wrapper.py (reported line 97)May include surrounding context.

python
return result
        else:
            # 实时输出模式
            process = subprocess.Popen(
                cmd,
                stdout=subprocess.PIPE,
                stderr=subprocess.STDOUT,

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · transcriber.py (reported line 31)May include surrounding context.

python
return result
        else:
            # 实时输出模式
            process = subprocess.Popen(
                cmd,
                stdout=subprocess.PIPE,
                stderr=subprocess.STDOUT,