Trust Assessment
whisper-mlx-local received a trust score of 10/100, placing it in the Untrusted category. This skill has significant security findings that require attention before use in production.
SkillShield's automated analysis identified 17 findings: 5 critical, 5 high, 6 medium, and 0 low severity. Key findings include Persistence / self-modification instructions, Network egress to untrusted endpoints, Unsafe environment variable passthrough.
The analysis covered 4 layers: Manifest Analysis, Static Code Analysis, Dependency Graph, LLM Behavioral Safety. The Manifest Analysis layer scored lowest at 0/100, indicating areas for improvement.
Last analyzed on February 13, 2026 (commit 13146e6a). SkillShield performs automated 4-layer security analysis on AI skills and MCP servers.
Layer Breakdown
Behavioral Risk Signals
Security Findings17
| Severity | Finding | Layer | Location | |
|---|---|---|---|---|
| CRITICAL | Persistence / self-modification instructions macOS LaunchAgent/LaunchDaemon persistence Remove any persistence mechanisms. Skills should not modify system startup configurations, crontabs, LaunchAgents, systemd services, or shell profiles. | Manifest | skills/impkind/whisper-mlx-local/SKILL.md:87 | |
| CRITICAL | Network egress to untrusted endpoints HTTP request to raw IP address Review all outbound network calls. Remove connections to webhook collectors, paste sites, and raw IP addresses. Legitimate API calls should use well-known service domains. | Manifest | skills/impkind/whisper-mlx-local/scripts/daemon.py:270 | |
| CRITICAL | Credential harvesting Reading well-known credential environment variables Skills should only access environment variables they explicitly need. Bulk environment dumps (os.environ.copy, JSON.stringify(process.env)) are almost always malicious. Remove access to Keychain, GPG keys, and credential stores. | Manifest | skills/impkind/whisper-mlx-local/scripts/transcriber.py:114 | |
| CRITICAL | Credential harvesting Reading well-known credential environment variables Skills should only access environment variables they explicitly need. Bulk environment dumps (os.environ.copy, JSON.stringify(process.env)) are almost always malicious. Remove access to Keychain, GPG keys, and credential stores. | Manifest | skills/impkind/whisper-mlx-local/scripts/transcriber.py:181 | |
| CRITICAL | Shell script allows Python code injection via filename The `scripts/transcribe_large.sh` script directly interpolates the `$AUDIO_FILE` variable into a Python heredoc string. An attacker can craft a malicious filename containing double quotes and Python code (e.g., `foo.ogg" + __import__('os').system('evil_command') + "`) to execute arbitrary commands on the system where the script runs. Do not directly embed user-controlled variables into code strings. Instead, pass the filename as a command-line argument to the Python script and parse it using `argparse`, or use a robust method for escaping strings if embedding is unavoidable. | LLM | scripts/transcribe_large.sh:39 | |
| HIGH | Unsafe environment variable passthrough Access to well-known credential environment variables Minimize environment variable exposure. Only pass required, non-sensitive variables to MCP servers. Use dedicated secret management instead of environment passthrough. | Manifest | skills/impkind/whisper-mlx-local/scripts/transcriber.py:114 | |
| HIGH | Unsafe environment variable passthrough Access to well-known credential environment variables Minimize environment variable exposure. Only pass required, non-sensitive variables to MCP servers. Use dedicated secret management instead of environment passthrough. | Manifest | skills/impkind/whisper-mlx-local/scripts/transcriber.py:181 | |
| HIGH | Persistence mechanism: macOS LaunchAgent Detected macOS LaunchAgent pattern. Persistence mechanisms allow malware to survive system restarts. Review this persistence pattern. Skills should not modify system startup configuration. | Static | skills/impkind/whisper-mlx-local/SKILL.md:87 | |
| HIGH | Daemon allows arbitrary file read via JSON path parameter The `scripts/daemon.py` HTTP server accepts POST requests to `/transcribe` with `Content-Type: application/json`. The JSON payload can contain a `file` or `path` key whose value is directly used as `audio_path`. The daemon then calls `os.path.exists()` and `transcriber.transcribe()` with this user-controlled path. This allows an attacker to check for the existence of arbitrary files and read their content, leading to data exfiltration. Do not allow direct file path specification from untrusted JSON input. Instead, require file uploads via multipart/form-data or stream the audio directly. If file paths must be used, implement strict validation to ensure paths are within an allowed directory and sanitize them thoroughly. | LLM | scripts/daemon.py:100 | |
| HIGH | Core `Transcriber` class performs arbitrary file read The `Transcriber` class in `scripts/transcriber.py` directly opens and reads the `audio_path` provided to its `transcribe` method for all supported backends (OpenAI, Groq, MLX, local). If an attacker can control the `audio_path` argument passed to this method (as demonstrated by the daemon vulnerability), they can read the content of any file accessible to the Python process. The `Transcriber` class should not directly accept arbitrary file paths from untrusted sources. It should either receive file-like objects (already opened and validated by a caller) or enforce strict path validation and sandboxing if it must handle paths directly. | LLM | scripts/transcriber.py:270 | |
| MEDIUM | Sensitive environment variable access: $HOME Access to sensitive environment variable '$HOME' detected in shell context. Verify this environment variable access is necessary and the value is not exfiltrated. | Static | skills/impkind/whisper-mlx-local/scripts/transcribe.sh:96 | |
| MEDIUM | Unpinned Python dependency version Requirement 'python-dotenv>=1.0.0' is not pinned to an exact version. Pin Python dependencies with '==<exact version>'. | Dependencies | skills/impkind/whisper-mlx-local/requirements.txt:5 | |
| MEDIUM | Unpinned Python dependency version Requirement 'openai>=1.12.0' is not pinned to an exact version. Pin Python dependencies with '==<exact version>'. | Dependencies | skills/impkind/whisper-mlx-local/requirements.txt:8 | |
| MEDIUM | Unpinned Python dependency version Requirement 'groq>=0.4.0' is not pinned to an exact version. Pin Python dependencies with '==<exact version>'. | Dependencies | skills/impkind/whisper-mlx-local/requirements.txt:11 | |
| MEDIUM | Unpinned Python dependency version Requirement 'faster-whisper>=1.0.0' is not pinned to an exact version. Pin Python dependencies with '==<exact version>'. | Dependencies | skills/impkind/whisper-mlx-local/requirements.txt:14 | |
| MEDIUM | Unpinned Python dependency version Requirement 'lightning-whisper-mlx>=0.0.10; sys_platform == "darwin" and platform_machine == "arm64"' is not pinned to an exact version. Pin Python dependencies with '==<exact version>'. | Dependencies | skills/impkind/whisper-mlx-local/requirements.txt:18 | |
| INFO | Skill loads environment variables from `.env` files The `scripts/transcriber.py` module uses `dotenv.load_dotenv()` to load environment variables. While a common practice, if the skill is deployed in an environment where an attacker can write to directories searched by `load_dotenv()` (e.g., the skill's installation directory or parent directories), they could place a malicious `.env` file to override legitimate credentials or inject other environment variables, potentially leading to credential harvesting or other attacks. Ensure that the skill's deployment environment restricts write access to directories where `.env` files might be loaded from. Consider explicitly specifying the path to a trusted `.env` file if possible, or relying solely on system-level environment variables for sensitive information. | LLM | scripts/transcriber.py:40 |
Scan History
Embed Code
[](https://skillshield.io/report/2c73b2cda009d299)
Powered by SkillShield