Back to skill

Security audit

Voice Transcriber Toolkit

Security checks for vulnerabilities and agentic risk

Overview

This voice transcription skill is mostly purpose-aligned, but it can return fabricated transcript text as a successful-looking result when a speech engine is unavailable.

Install only if you are comfortable with local audio files being processed by Python speech libraries and ffmpeg, and do not use this skill for compliance, evidence, accessibility, or other high-reliability transcript workflows unless the mock fallback is removed or your agent checks the returned note/model/status before trusting transcript text. Pin dependencies or use an isolated environment before installing.

Vulnerability Patterns
  • 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
  • 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)

T08 · Insecure Dependencies

Warning
Location
requirements.txt:1
Finding

Mutable and Unverified Python Dependencies

Content
View full analysis
=20230314 vosk>=0.3.45 pytest>=7.0.0 ``` Associated installation configuration: ```yaml install: - id: python-deps kind: python requirements: "requirements.txt" - id: ffmpeg kind: apt packages: ["ffmpeg"] ``` ### Technical Analysis All Python dependencies use open-ended minimum-version constraints. The installer can therefore resolve and install future package versions that were not present during the audit. No lock file, exact transitive dependency versions, artifact hashes, or trusted-index restrictions are supplied. Python package installation can execute package build or installation logic. Consequently, compromise of an allowed future package release, one of its transitive dependencies, or the configured package repository could result in attacker-controlled code executing during installation or subsequent import. The test framework `pytest` is included in the primary requirements file even though it is only needed for development and testing. This unnecessarily expands the production dependency graph and supply-chain attack surface. ### Attack Path 1. An attacker compromises a permitted future release of a listed package, one of its transitive dependencies, or the package source used by the installer. 2. The Skill is installed or its dependencies are upgraded after the compromised version becomes eligible under the `>=` constraints. 3. The package resolver selects and downloads the compromised artifact because no exact version or hash prevents it. 4. Attacker-controlled installation, build, or imported runtime code executes. 5. The payload gains the same local permissions and environmental access as the user or service performing installation or running the Sk ...[truncated 595 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
scripts/voice_transcriber.py:25
Finding

Fabricated Transcription Returned When a Recognition Engine Is Unavailable

Content
View full analysis
Dict: import random sample_texts = { "en": "This is a sample transcription. The quick brown fox jumps over the lazy dog.", # Additional built-in sample translations are defined here. } lang = language or random.choice(list(sample_texts.keys())) text = sample_texts.get(lang, sample_texts["en"]) return { "text": text, "language": lang, "segments": [ {"id": 0, "start": 0.0, "end": 2.5, "text": text[:30]}, {"id": 1, "start": 2.5, "end": 5.0, "text": text[30:]}, ], "duration": 5.0, "engine": self.engine, "model": f"{self.model_size} (mock)", "note": "Mock result - install whisper/vosk for real transcription", } ``` ### Technical Analysis Failure to import a requested recognition engine does not cause initialization or transcription to fail. Instead, the object remains usable with `_model` set to `None`, and every subsequent call returns predefined text unrelated to the supplied audio. Althoug ...[truncated 1842 chars]
Remediation
View remediation
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (12)

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
70% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · README.md (reported line 68)May include surrounding context.

bash
pip install -r requirements.txt
# 同时需要安装 ffmpeg:
# Ubuntu/Debian: sudo apt install ffmpeg
# macOS: brew install ffmpeg

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill declares installation steps and demonstrates functionality that relies on shell execution and writing output files, but it does not declare any explicit tool scope such as permissions or allowed-tools. This creates an authorization and review gap: a host may grant broader-than-necessary capabilities or users may be unaware that the skill can invoke ffmpeg/python and write files, increasing the risk of unintended command execution or file modification if the implementation is unsafe.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The Vosk model is initialized with Model(lang="en"), which hard-codes English as the language/locale. This is a natural-language policy concern because the skill supports multiple languages elsewhere but silently constrains Vosk users to English without explicit opt-in or clear documentation of that limitation.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

When the model is unavailable, transcribe() silently returns fabricated text that looks like a legitimate transcription result. In security-sensitive or compliance-sensitive workflows, this can cause false records, bad decisions, integrity failures, or downstream automation to trust invented content as if it came from the audio.

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 · scripts/voice_transcriber.py (reported line 223)May include surrounding context.

python
def _check_ffmpeg(self) -> bool:
        import subprocess
        try:
            subprocess.run(["ffmpeg", "-version"], capture_output=True, check=True)
            return True
        except (subprocess.CalledProcessError, FileNotFoundError):
            return False

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 · scripts/voice_transcriber.py (reported line 250)May include surrounding context.

python
"-y", output_path
        ]

        result = subprocess.run(cmd, capture_output=True, text=True)
        if result.returncode != 0:
            raise RuntimeError(f"ffmpeg conversion failed: {result.stderr}")

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 · scripts/voice_transcriber.py (reported line 266)May include surrounding context.

python
]

        try:
            result = subprocess.run(cmd, capture_output=True, text=True, check=True)
            data = json.loads(result.stdout)

            format_info = data.get("format", {})

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The example call hard-codes language="zh", which suggests a default or expected language behavior in the skill documentation. Because the file does not mention that users may choose another language or that the tool is intentionally region-specific, this can conflict with locale-choice policy expectations.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
95% confidence
Finding

The dependency specification for openai-whisper uses a lower-bound range rather than an exact pinned version, which makes builds non-reproducible and can pull in newly released packages with unexpected vulnerabilities or breaking changes. In a security-sensitive supply-chain context, this increases risk even though it is common and likely not intentional.

Content

Scanner excerpt · requirements.txt (reported line 1)May include surrounding context.

text
openai-whisper>=20230314
vosk>=0.3.45
pytest>=7.0.0

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
95% confidence
Finding

The dependency specification for vosk is unpinned (>=0.3.45), allowing installation of any later release. This weakens reproducibility and increases supply-chain exposure because future versions may introduce vulnerabilities, malicious compromise, or incompatible behavior without explicit review.

Content

Scanner excerpt · requirements.txt (reported line 2)May include surrounding context.

text
openai-whisper>=20230314
vosk>=0.3.45
pytest>=7.0.0

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
98% confidence
Finding

The pytest dependency is also unpinned, so installations may resolve to different versions over time. Even though pytest is primarily a development/test dependency, leaving it unpinned can still expose the environment to known or newly introduced issues and makes it harder to verify whether a safe version is installed.

Content

Scanner excerpt · requirements.txt (reported line 3)May include surrounding context.

text
openai-whisper>=20230314
vosk>=0.3.45
pytest>=7.0.0

Unverifiable Dependency: pytest has 2 known advisory(ies) (CVE-2025-71176 (pytest has vulnerable tmpdir handling); CVE-2025-71176 (pytest has vulnerable tmpdir handling)), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low
Category
Supply Chain
Confidence
89% confidence
Finding

The manifest includes pytest without pinning a version, and the analyzer indicates there are known advisories affecting some pytest releases. Because the version is not fixed, it is impossible to determine from this file alone whether installation will select a vulnerable release, creating uncertainty and potential exposure in development or CI environments.

Content

No source excerpt is available for this finding.