Security audit

llm-provider Whisper

Security checks for vulnerabilities and agentic risk

Overview

The skill appears to be a low-quality Whisper transcription skill, but its instructions are inconsistent and request broad read/write/command authority plus API-key handling beyond the stated local transcription purpose.

Review before installing. Use this only if you are comfortable with an agent reading and writing files and running local commands, and verify the actual Whisper dependency and authentication requirements yourself. Do not provide API keys unless the publisher clarifies why a local Whisper transcription skill needs them.

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 (4)

Intent-Code Divergence

Medium

Confidence: 96% confidence
Finding: The skill documentation makes contradictory security-relevant claims about authentication, advertising 'no API key' use while later requiring an API key and showing export instructions. This can mislead users and agents into selecting or configuring the skill under false assumptions, increasing the chance of unsafe credential handling, failed execution paths, or invoking an unintended remote service.

Intent-Code Divergence

Medium

Confidence: 94% confidence
Finding: The FAQ claims no extra software installation is needed, but the skill is described as depending on a local Whisper CLI command-line tool. This mismatch can cause users or agents to execute workflows in environments missing required binaries, leading to insecure workarounds, ad hoc installation steps, or invoking alternative tools not covered by the skill's security assumptions.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The description uses broad activation language such as recommending the skill for API integration, webhooks, and system connections, even though the skill is ostensibly a local speech-to-text tool. Overbroad invocation guidance can cause an agent to select this skill in unrelated contexts where it may read files, execute commands, or handle credentials unnecessarily, expanding attack surface and increasing the risk of misuse.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The feature and scenario sections are generic enough to fit many unrelated automation tasks, rather than clearly constraining the skill to transcription workflows. In an agent ecosystem, vague scope definitions increase the chance of accidental invocation with sensitive inputs or privileged operations, especially because the declared toolset includes read, write, and exec capabilities.

Static analysis

No suspicious patterns detected.