ClawHub
Back to skill

Security audit

音频流上传免费版

Security checks for vulnerabilities and agentic risk

Overview

This looks like a real audio-upload helper, but its trigger scope is broader than its actual upload purpose and it can send local media plus API credentials to a third-party service.

Review before installing. Use this only for explicit audio upload tasks, only with files you intentionally choose, and only if you trust the third-party streaming API. Store API keys in environment variables or a secret manager, avoid pasting real keys into shell commands or logs, and do not rely on this skill for general video processing, editing, conversion, or dubbing workflows.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)

Description-Behavior Mismatch

Medium
Confidence
93% confidence
Finding
The skill is presented as a narrowly scoped audio upload utility, but the documentation expands its apparent capabilities to generic create/query/modify/delete, import/export, save, and convert operations. This kind of scope inflation can cause an agent to invoke the skill for requests outside its intended trust boundary, increasing the chance of unintended actions or misuse.

Intent-Code Divergence

Medium
Confidence
90% confidence
Finding
The skill warns against hardcoding secrets, yet its executable examples place secret-bearing headers directly into curl commands and scripts. Even when placeholders or environment variables are used, this pattern encourages exposing credentials in command history, process listings, logs, or copied scripts, undermining the stated secret-handling guidance.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The trigger condition is overly broad, covering generic video processing, audio editing, media conversion, and dubbing-generation use cases that exceed the skill's actual upload-focused function. This increases the likelihood that an agent will invoke the skill in inappropriate contexts and transmit files or credentials to an external service when the user did not specifically request uploading.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The top-level description uses broad, vague usage guidance that overlaps with many common media tasks unrelated to simple audio upload. In an agent environment, ambiguous invocation guidance can route unrelated user requests into a skill that performs external network actions and handles credentials, creating unnecessary exposure.

External Transmission

Medium
Category
Data Exfiltration
Content
}

# 步骤1：创建音频对象
create_resp = requests.post(f'{BASE_URL}/videos/create', headers=HEADERS, json={
    'title': '我的原创音乐',
    'type': 'audio'
})
Confidence
95% confidence
Finding
The example performs an outbound HTTP request to a third-party domain and sends user content plus API credentials in headers. External transmission is expected for an upload skill, but it is still security-relevant because an agent may exfiltrate local files or sensitive metadata to a non-platform-controlled service if invoked too broadly or without clear user consent.

External Transmission

Medium
Category
Data Exfiltration
Content
END_POS=$((FILE_SIZE - 1))
HASH=$(md5sum /path/to/podcast.mp3 | awk '{print $1}')

curl -s -X POST "https://api-w3stream.attoaioz.cyou/api/videos/AUDIO_ID/part" \
  -H 'stream-public-key: YOUR_PUBLIC_KEY' \
  -H 'stream-secret-key: YOUR_SECRET_KEY' \
  -H "Content-Range: bytes 0-$END_POS/$FILE_SIZE" \
Confidence
95% confidence
Finding
This curl example uploads a local file to an external domain while transmitting API credentials and file metadata. In the context of an agent skill with exec capability and broad triggers, this creates a real risk of unintended data exfiltration or secret disclosure if the skill is invoked for loosely related media tasks.

Static analysis

No suspicious patterns detected.