Security audit

Meixin Meeting Assistant

Security checks for vulnerabilities and agentic risk

Overview

This skill mostly does what it claims, but it embeds a default cloud API key and handles sensitive internal meeting recordings through external transcription and persistent storage without enough safeguards.

Review before installing. Do not use this with confidential, regulated, or internal recordings unless external DashScope processing and public or CDN-hosted audio URLs are approved. Remove and rotate the embedded API key, require your own secure DASHSCOPE_API_KEY, and define how raw transcripts and minutes will be retained or deleted.

SkillSpector

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 (9)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 94% confidence
Finding: The documented behavior materially differs from what is implemented: the code sends audio to an external service, depends on a public URL/CDN workflow not disclosed to users, and includes a built-in API key. This mismatch prevents informed consent and makes it easier for sensitive meeting audio and metadata to be exfiltrated or mishandled under the guise of a benign note-taking skill.

Context-Inappropriate Capability

High

Confidence: 99% confidence
Finding: The skill embeds a concrete default `DASHSCOPE_API_KEY` directly in documentation code, which is a credential exposure issue regardless of whether the key is still valid. Hardcoded secrets can be abused for unauthorized API usage, billing fraud, and unauthorized access to external AI services, and they also normalize insecure secret-handling practices.

Context-Inappropriate Capability

High

Confidence: 99% confidence
Finding: The script embeds a live-looking DashScope API key directly in source code as a default value. Hardcoded secrets are highly exposed through source control, logs, packaging, and reuse by unauthorized parties, enabling abuse of the account and potentially access to associated transcription data or billable API usage. In a meeting-transcription skill handling internal recordings, this is especially dangerous because it processes sensitive corporate audio.

Missing User Warnings

High

Confidence: 96% confidence
Finding: The skill description fails to warn users that meeting audio and related content will be transmitted to an external ASR provider. In the context of internal health-insurance meetings, this omission is especially dangerous because recordings may contain confidential business data, personal data, and regulated health-related information.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The skill omits a user-facing warning that raw transcripts and generated minutes are written to persistent workspace memory files. This creates a confidentiality risk because sensitive meeting content and inferred speaker identities may remain accessible long after the task completes.

Missing User Warnings

High

Confidence: 97% confidence
Finding: Using a default embedded API key means the script will silently access an external service under the author's credential without explicit operator awareness or consent. This weakens accountability, can mask unauthorized third-party data transfer, and exposes the credential to theft and misuse. Given the skill's role in transcribing internal meeting audio, the privacy and financial consequences are materially elevated.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The script uploads or references meeting audio to a third-party ASR provider and submits it for transcription without presenting a clear privacy warning or confirming authorization to share the content externally. Internal meeting recordings commonly contain confidential business, employee, customer, or health-insurance-related information, so undisclosed transmission to a cloud vendor creates a real privacy and compliance risk.

Ssd 3

Medium

Confidence: 92% confidence
Finding: The instructions require recording all speech and archiving transcripts/minutes in persistent memory, including attributed speakers. For internal meetings, this default-all retention model increases exposure of confidential business discussions and personal data, especially when speaker identification is inferred from prior records.

Ssd 3

Medium

Confidence: 90% confidence
Finding: The workflow and quality rules instruct the system to preserve all utterances and store both transcripts and minutes, which is risky for potentially sensitive meeting material. In this business context, full-fidelity retention can expose strategy, personnel information, insurance operations details, and other confidential content by default.

Static analysis

No suspicious patterns detected.