Back to skill

Security audit

SenseAudio Voice CN

Security checks for vulnerabilities and agentic risk

Overview

This voice skill uses cloud speech services and local audio files in ways that match its stated purpose, with some setup and privacy caveats users should understand.

Install only if you are comfortable sending text/audio to SenseAudio or Microsoft Edge TTS and storing generated files locally. Prefer a dedicated SENSE_API_KEY, avoid sensitive recordings, review the extra OpenClaw config lookup, and install dependencies in a virtual environment with pinned versions when possible.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T08 · Insecure Dependencies

Warning
Location
scripts/setup.sh:12
Finding
Unpinned Python dependency installation permits supply-chain substitution## Vulnerability Details **File Location**: `scripts/setup.sh`, lines 12-15 **Vulnerability Type**: Unpinned third-party dependency installation **Risk Level**: Medium ### Vulnerable Code ```bash # Install requests library (the only dependency) echo "📦 Installing Python dependencies..." pip3 install requests --quiet echo " ✅ requests installed" ``` ### Technical Analysis The setup script installs `requests` directly from the Python package index without an exact version, integrity hash, lockfile, isolated virtual environment, or explicit trusted index. Consequently, installation results can change over time without any modification to the reviewed Skill. Although `requests` is a legitimate and widely used dependency, the command resolves its latest compatible release and its transitive dependencies at installation time. If the package publisher, package index, configured mirror, DNS/proxy infrastructure, or a transitive dependency is compromised, a malicious distribution could be selected. Python package installation can execute build-backend code while processing source distributions. A malicious package can therefore execute code during `pip3 install`, before the Skill imports or invokes the dependency. The absence of a virtual environment also allows the command to modify the invoking user's global or user-level Python environment. The same supply-chain concern applies to the documentation's unpinned `pip install edge-tts` instruction, but the directly executable installation sink is the command shown above. ### Attack Path 1. An attacker compromises the `requests` publishing account, a transitive dependency, the configured Python index or mirror, or another component in the package-resolution path. 2. The attacker publishes or serves a malicious package version or source distribution. 3. A user executes `scripts/setup.sh`. 4. `pip3 install requests --quiet` resolves the attacker-controlled artifact because ...[truncated 1296 chars]
Remediation
## Remediation Suggestions 1. Pin every direct and transitive Python dependency to an reviewed version in a dedicated requirements or lock file. 2. Record cryptographic hashes and enforce them during installation: ```text requests==2.32.5 \ --hash=sha256:<verified-wheel-hash> ``` ```bash python3 -m pip install \ --require-hashes \ --only-binary=:all: \ -r requirements.txt ``` 3. Generate and review a fully resolved lock file with tooling such as `pip-compile`, including hashes for all transitive dependencies. 4. Install dependencies into a Skill-specific virtual environment rather than the global Python environment: ```bash python3 -m venv .venv .venv/bin/python -m pip install --require-hashes -r requirements.txt ``` 5. Use `python3 -m pip` instead of a standalone `pip3` command to ensure dependencies are installed for the intended interpreter. 6. Explicitly configure the approved package index and prevent fallback to untrusted extra indexes. 7. Remove `--quiet` during security-sensitive installation or emit the resolved versions and artifact sources so users can audit what was installed. 8. Pin `edge-tts` and its transitive dependencies in the same lock file rather than directing users to run an unpinned installation command. 9. Add automated dependency scanning and a controlled update process in which version and hash changes are reviewed before release.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (37)

Tainted flow: 'headers' from os.environ.get (line 58, credential/environment) → requests.post (network output)

Critical
Category
Data Flow
Content
data["target_language"] = translate_to
    
    try:
        response = requests.post(DEFAULT_ASR_URL, headers=headers, files=files, data=data, timeout=120)
        response.raise_for_status()
        
        result = response.json()
Confidence
90% confidence
Finding
Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Tainted flow: 'headers' from os.environ.get (line 283, credential/environment) → requests.post (network output)

Critical
Category
Data Flow
Content
}
    
    try:
        response = requests.post(DEFAULT_TTS_URL, headers=headers, json=payload, timeout=timeout)
        response.raise_for_status()
        
        data = response.json()
Confidence
90% confidence
Finding
Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding
The skill description omits or understates several security-relevant behaviors: creating/storing audio files locally, invoking local system players, reading credentials from config/environment, and using third-party services beyond the named provider. These undeclared behaviors reduce transparency and informed consent, and can lead to credential exposure, unexpected persistence of sensitive audio/text, or execution of local commands in environments where such actions are restricted.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding
The skill description omits or understates several security-relevant behaviors: creating/storing audio files locally, invoking local system players, reading credentials from config/environment, and using third-party services beyond the named provider. These undeclared behaviors reduce transparency and informed consent, and can lead to credential exposure, unexpected persistence of sensitive audio/text, or execution of local commands in environments where such actions are restricted.

Tainted flow: 'files' from open (line 63, file read) → requests.post (network output)

High
Category
Data Flow
Content
data["target_language"] = translate_to
    
    try:
        response = requests.post(DEFAULT_ASR_URL, headers=headers, files=files, data=data, timeout=120)
        response.raise_for_status()
        
        result = response.json()
Confidence
80% confidence
Finding
File contents flow to a network sink. This may indicate data exfiltration of sensitive files.

Natural-Language Policy Violations

Medium
Confidence
78% confidence
Finding
The documentation presents fixed language-to-engine routing that automatically sends Chinese text to SenseAudio and English/Japanese text to Edge TTS, without emphasizing informed user choice or privacy implications of provider selection. This can cause users to transmit content to a service they did not intend to use, especially when language detection is automatic.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The README instructs users to store a sensitive API key and to send TTS content to remote services, but it does not clearly warn that text and credentials are being transmitted to third-party providers. In a voice skill, users may reasonably assume local processing unless documentation explicitly states otherwise, so this omission creates privacy and secret-handling risk.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
python tts.py --check-players

# 安装推荐播放器
sudo apt-get install alsa-utils    # aplay (WAV 原生支持)
sudo apt-get install pulseaudio    # paplay
sudo apt-get install ffmpeg        # ffplay (多格式支持)
```
Confidence
70% confidence
Finding
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
python tts.py --check-players

# 安装推荐播放器
sudo apt-get install alsa-utils    # aplay (WAV 原生支持)
sudo apt-get install pulseaudio    # paplay
sudo apt-get install ffmpeg        # ffplay (多格式支持)
```
Confidence
70% confidence
Finding
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
python tts.py --check-players

# 安装推荐播放器
sudo apt-get install alsa-utils    # aplay (WAV 原生支持)
sudo apt-get install pulseaudio    # paplay
sudo apt-get install ffmpeg        # ffplay (多格式支持)
```
Confidence
70% confidence
Finding
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill advertises behavior that requires network access, local file read/write, environment/config access, and shelling out to local audio players, but it declares no explicit tool scope or permissions. In an agent environment, this weakens isolation and informed consent because the runtime may grant broader capabilities than users expect, enabling unintended data access or command execution paths.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The documentation promotes automatic playback and local storage without clearly warning that user-supplied text and generated audio may be transmitted to third-party services and persisted on disk. In a voice skill, prompts may contain sensitive personal, workplace, or educational content, so insufficient disclosure can lead to unintentional data leakage and retention.

External Transmission

Medium
Category
Data Exfiltration
Content
import requests

# 默认配置
DEFAULT_ASR_URL = "https://api.senseaudio.cn/v1/audio/transcriptions"
DEFAULT_ASR_MODEL = "sense-asr-deepthink"
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
import requests

# 默认配置
DEFAULT_ASR_URL = "https://api.senseaudio.cn/v1/audio/transcriptions"
DEFAULT_ASR_MODEL = "sense-asr-deepthink"
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
import requests

# 默认配置
DEFAULT_ASR_URL = "https://api.senseaudio.cn/v1/audio/transcriptions"
DEFAULT_ASR_MODEL = "sense-asr-deepthink"
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
import requests

# 默认配置
DEFAULT_ASR_URL = "https://api.senseaudio.cn/v1/audio/transcriptions"
DEFAULT_ASR_MODEL = "sense-asr-deepthink"
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
import requests

# 默认配置
DEFAULT_ASR_URL = "https://api.senseaudio.cn/v1/audio/transcriptions"
DEFAULT_ASR_MODEL = "sense-asr-deepthink"
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Context-Inappropriate Capability

Medium
Confidence
90% confidence
Finding
The script reads credentials not only from the expected environment variable but also from unrelated local OpenClaw config paths, including a specific agent directory. This broadens secret sourcing in a surprising way and can cause accidental credential use or cross-skill secret exposure if another local config contains a valid API key.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The script uploads potentially sensitive audio content to a third-party cloud ASR service without an explicit privacy warning or affirmative consent step at runtime. In a skill context, users may treat a local script as on-device processing and unintentionally transmit personal, confidential, or regulated voice data externally.

Natural-Language Policy Violations

Medium
Confidence
98% confidence
Finding
The file's comments, status messages, and example usage are written in Chinese throughout, which effectively forces a single language for users interacting with the skill. Under the policy, locale or language constraints should either be optional for the user or clearly documented as a justified regional restriction.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
echo "   ✅ aplay 已安装"
else
    echo "   ⚠️  未找到音频播放器"
    echo "   建议安装:sudo apt-get install ffmpeg"
fi

# 验证 API Key
Confidence
70% confidence
Finding
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
echo "   ✅ aplay 已安装"
else
    echo "   ⚠️  未找到音频播放器"
    echo "   建议安装:sudo apt-get install ffmpeg"
fi

# 验证 API Key
Confidence
70% confidence
Finding
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
echo "   ✅ aplay 已安装"
else
    echo "   ⚠️  未找到音频播放器"
    echo "   建议安装:sudo apt-get install ffmpeg"
fi

# 验证 API Key
Confidence
70% confidence
Finding
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
echo "   ✅ aplay 已安装"
else
    echo "   ⚠️  未找到音频播放器"
    echo "   建议安装:sudo apt-get install ffmpeg"
fi

# 验证 API Key
Confidence
70% confidence
Finding
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
echo "   ✅ aplay 已安装"
else
    echo "   ⚠️  未找到音频播放器"
    echo "   建议安装:sudo apt-get install ffmpeg"
fi

# 验证 API Key
Confidence
70% confidence
Finding
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Static analysis

No suspicious patterns detected.