Back to skill

Security audit

Audio Transcriber

Security checks for vulnerabilities and agentic risk

Overview

The skill is clearly for audio transcription, but it asks agents to record/process microphone audio without enough consent, retention, or scoping guidance.

Review this before installing if you may use it for meetings or live microphone input. Only run recording commands after all participants consent, use short durations, know where WAV/transcript outputs are stored, and prefer a pinned, reviewed dependency environment.

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
SKILL.md:118
Finding

Unpinned Third-Party Dependency Installation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 118
Vulnerability Type: Unpinned and unverifiable third-party dependencies
Risk Level: Medium

Vulnerable Code:

bash
pip install openai-whisper pyaudio numpy

Technical Analysis

The skill instructs users or agents to install three packages without exact version constraints, package hashes, a lockfile, or an explicitly trusted package index. Consequently, installation results can change over time and may include unreviewed versions or transitive dependencies.

Because Python packages can execute code during installation and when imported, compromise of a named package, its maintainer account, the package repository, or a transitive dependency could result in attacker-controlled code execution. The reviewed project does not itself prove that any named package is malicious; the vulnerability is the unsafe, non-reproducible dependency installation process.

Attack Path

  1. An attacker compromises a listed package, its release process, or one of its transitive dependencies and publishes a malicious version to the package source used by pip.
  2. A user or agent follows the instruction at line 118.
  3. Because no versions or hashes are specified, pip resolves the attacker-controlled or otherwise unreviewed release.
  4. Malicious package code executes during installation or later when the dependency is imported by the external audio implementation.
  5. The payload operates with the privileges of the account running pip or the audio process.

Impact Assessment

Successful exploitation could permit arbitrary code execution with the installing user's privileges. Depending on those privileges and the host environment, the attacker could access files and environment variables available to that user, alter user-owned data, interfere with audio processing, or establish additional malicious behavior. The affected scope is the installation environmen ...[truncated 167 chars]

Remediation
View remediation

Remediation Suggestions

  1. Replace the direct installation command with a reviewed dependency manifest containing exact versions.
  2. Lock all transitive dependencies to make installations reproducible.
  3. Require package hashes, such as through a hash-locked requirements.txt installed with:
    bash
    python -m pip install --require-hashes -r requirements.txt
    
  4. Configure an approved HTTPS package index or internal artifact repository rather than relying on an implicit source.
  5. Review and scan every pinned package and transitive dependency before release.
  6. Perform installation in an isolated virtual environment under a non-privileged account.
  7. Periodically update dependencies through a controlled review and testing process rather than allowing automatic resolution to arbitrary new releases.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill documentation describes recording, transcription, translation, and emotion analysis but provides no privacy notice, consent guidance, retention behavior, or warning that captured audio may contain sensitive personal or third-party information. This is especially risky for an audio-processing skill because users may unknowingly record confidential conversations or biometric-like voice data without informed consent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger description is broad enough that the skill may be auto-loaded for loosely related requests such as meetings, translation, or audio analysis without clear user intent to record or process audio. In this context, unintended activation is more dangerous because the skill handles microphone/audio data, which can expose sensitive conversations or cause privacy-invasive recording workflows to start unexpectedly.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

文档主体与使用说明均以中文固定呈现,且未说明是否支持按用户偏好切换语言。按规则,若技能对语言/locale 有强制性约束而没有用户选择或明确合理的地区限定,属于自然语言策略风险。

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.