Security audit

ai-podcast-free

Security checks for vulnerabilities and agentic risk

Overview

The skill’s core text-to-podcast workflow is understandable, but its broad triggers and overstated file/translation capabilities could cause users or agents to send the wrong content to an external service.

Review before installing. Use this only for text you intentionally want to send to MagicPodcast for podcast generation. Do not rely on it for translation, localization, PDFs, arbitrary file processing, or custom voice controls unless the publisher updates the documentation and API fields to clearly support those features.

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The manifest and usage guidance advertise translation, multilingual conversion, and localization use cases, but the documented implementation only submits text and a language field to a podcast-generation API. This mismatch can cause agents or users to route unrelated or sensitive translation tasks to the skill under false assumptions, leading to unintended external data disclosure and unsafe task delegation.

Intent-Code Divergence

Medium

Confidence: 93% confidence
Finding: The FAQ claims support for custom host voice styles, but the documented API call only includes text and language and shows no voice-style parameter. This can mislead downstream agents into collecting extra user preferences or making incorrect promises, reducing trust and potentially causing unnecessary transmission of user data to an external service.

Intent-Code Divergence

Medium

Confidence: 96% confidence
Finding: The comparison section claims PDF-to-podcast capability in the free version, directly contradicting the stated capability boundary that PDF input is not supported. This inconsistency can cause agents to select the skill for unsupported document-processing tasks, leading to failed executions, mishandling of user files, or accidental forwarding of inappropriate content to external tooling.

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The documentation advertises file read/parse/write capabilities that are inconsistent with the described text-only podcast workflow. In an agent environment with read/write/exec tools enabled, overstating these capabilities can prompt broader file access than necessary and increase the chance of inappropriate handling of local data.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The trigger description is overly broad and ambiguous, including unrelated use cases like translation and localization. Broad routing criteria increase the likelihood that an agent will invoke the skill for tasks outside its real scope, which is especially risky here because the skill sends user content to an external API.

Static analysis

No suspicious patterns detected.