ClawHub
Back to skill

Security audit

Weryai Podcast Generator

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed WeryAI podcast API wrapper; the flagged upload and music-generation code is extra vendored code that is not used by the documented podcast entrypoints.

Install only if you trust this package with a WeryAI API key and paid account credits. Keep WERYAI_BASE_URL on the official WeryAI API host unless you intentionally trust another endpoint, use dry-run before paid generation, and stick to the documented podcast scripts rather than the extra vendored music helpers.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (3)

Context-Inappropriate Capability

High
Confidence
98% confidence
Finding
The helper uploads local files to whatever ctx.baseUrl points to and only emits a warning for non-official domains, which means sensitive local files can be exfiltrated to attacker-controlled endpoints if configuration is influenced upstream. In an agent skill context, this is especially dangerous because skills often bridge user-controlled inputs and local filesystem access, turning a misconfiguration or prompt-driven endpoint change into data exfiltration.

Description-Behavior Mismatch

High
Confidence
94% confidence
Finding
This file implements music generation and reference-audio upload logic inside a skill advertised as a podcast generator, creating a clear scope mismatch between declared capability and actual behavior. That mismatch is dangerous because it can cause users, reviewers, or higher-level orchestration to grant permissions or trust assumptions appropriate for podcast generation while the code performs materially different media-processing actions, including uploading user-provided audio to a remote API.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The code automatically resolves and uploads a local or provided reference audio source via resolvePublicUrlFromSource before submitting the generation request, with no evidence in this file of an explicit user-facing warning, consent check, or disclosure. In an agent setting, silent transmission of user audio or local file content to a third-party endpoint can expose sensitive data and violate user expectations, especially because dry-run messaging acknowledges the upload path only for previews rather than enforcing consent in real execution.

VirusTotal

67/67 vendors flagged this skill as clean.

View on VirusTotal