Security audit

音频转文字

Security checks for vulnerabilities and agentic risk

Overview

This skill does what it says: it transcribes user-selected audio with Baidu Cloud and writes local transcript outputs, with privacy and dependency cautions to consider.

Install only if you are comfortable sending the selected audio files to Baidu Cloud for transcription. Do not use it for confidential or regulated recordings unless your policy allows that transfer, prefer environment variables or a secrets manager for Baidu credentials, and store transcript/JSON outputs in a private directory. Pin or audit the requests dependency before enterprise use.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (11)

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The README describes using Baidu Cloud speech recognition but does not clearly disclose that users' audio content is transmitted to a third-party cloud provider for processing. This creates a real privacy and compliance risk because users may unknowingly upload sensitive recordings, and in enterprise or regulated environments that omission can lead to unauthorized data externalization.

Missing User Warnings

Low

Confidence: 79% confidence
Finding: The README mentions automatic saving of transcripts, JSON responses, and reports, but it does not prominently warn that multiple local output files containing potentially sensitive recognized text and metadata will be created. This can expose private content to other local users, backups, sync tools, or accidental sharing if operators do not realize the data is being persistently stored.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill documentation states it uses Baidu Cloud speech recognition but does not explicitly warn that user audio will be transmitted to a third-party cloud service. This can cause users to submit sensitive voice content without informed consent, creating privacy, compliance, and data-handling risk.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill instructs users to provide API Key and Secret Key via CLI, config files, and environment variables, but does not warn that these credentials are sensitive secrets. Users may expose them through shell history, committed config files, logs, screenshots, or shared environments, enabling unauthorized API use and possible billing abuse.

Missing User Warnings

Low

Confidence: 89% confidence
Finding: The documentation describes writing transcripts, raw JSON responses, and processing reports to disk without clearly warning that these files may contain sensitive spoken content and metadata. Users may unintentionally persist private information in accessible project directories, backups, or shared workspaces.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill base64-encodes audio content and sends it to Baidu's remote speech-recognition API, but there is no explicit user-facing disclosure, consent prompt, or privacy gate before off-device transfer. If users process sensitive recordings, private audio may be transmitted to a third party unexpectedly, creating confidentiality and compliance risks.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: In the long-audio path, each audio segment is transmitted to an external recognizer service without any explicit consent, disclosure, or policy gate in this module. That creates a privacy and data-handling risk because potentially sensitive voice content may leave the local environment unexpectedly.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The short-audio path sends the audio file to an external recognizer with no visible user-facing warning or consent mechanism. If the skill is used on confidential recordings, this can result in unintended disclosure of sensitive personal or business information.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The manifest states that the skill uses Baidu Cloud speech recognition and requires API credentials, but it does not clearly disclose that users' audio files will be uploaded to an external third-party service for processing. This creates a privacy and consent problem because audio may contain sensitive personal, business, or regulated information, and users may provide it without understanding the data transfer and retention implications.

Unpinned Dependencies

Low

Category: Supply Chain
Content: requests>=2.25.0
Confidence: 91% confidence
Finding: The dependency specifier `requests>=2.25.0` is unpinned, so installs can resolve to different versions over time and across environments. This weakens reproducibility and supply-chain control, and it may also allow resolution to an outdated vulnerable version such as 2.25.0 if a lockfile or resolver behavior permits it.

Known Vulnerable Dependency: requests==2.25.0 — 8 advisory(ies): CVE-2024-47081 (Requests vulnerable to .netrc credentials leak via malicious URLs); CVE-2024-35195 (Requests `Session` object does not verify requests after making first request wi); CVE-2026-25645 (Requests has Insecure Temp File Reuse in its extract_zipped_paths() utility func) +5 more

High

Category: Supply Chain
Confidence: 86% confidence
Finding: Although the file does not explicitly pin `requests==2.25.0`, the declared range `requests>=2.25.0` includes version 2.25.0 and other potentially affected releases, so the environment could resolve to a version with known advisories. If an application using this skill installs a vulnerable Requests release, issues such as credential leakage or TLS verification problems could expose secrets or enable man-in-the-middle attacks.

Static analysis

No suspicious patterns detected.