Security audit

Prd Review

Security checks across malware telemetry and agentic risk

Overview

This PRD review skill has a coherent purpose, but it embeds Feishu app credentials and automatically stores/reuses review data in ways users should review before installing.

Review before installing. Do not use the embedded Feishu credentials; they should be removed and rotated. Use this only with documents you are authorized to process, and require explicit confirmation before reading Feishu links, writing reports back, saving audit history, or reusing prior records.

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 (5)

Description-Behavior Mismatch

Medium

Confidence: 89% confidence
Finding: The skill persists audit-history JSON files locally for later reuse, but this retention is not surfaced in the top-level description or accompanied by consent/retention controls. Review artifacts may contain sensitive product plans, internal findings, and document metadata, creating an unnecessary data retention and secondary exposure risk.

Context-Inappropriate Capability

High

Confidence: 99% confidence
Finding: The skill embeds what appear to be live Feishu app credentials and instructs subprocess execution of an external MCP binary using those secrets. Hardcoded secrets can be extracted and abused for unauthorized document access or API use, and subprocess-based execution enlarges the trust boundary for a document-review skill that should not need embedded credentials in its prompt file.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The skill states that Feishu documents will be read automatically but does not provide a clear user-facing warning that external content access will occur. In a security-sensitive context, silent external reads can expose internal documents without informed consent and make accidental overcollection more likely.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill retains audit results locally in audit-history without warning users in the main description or flow summary. Hidden local persistence increases privacy and compliance risk because users may assume the review is ephemeral while the system keeps reusable records of sensitive PRD-derived information.

Ssd 3

Medium

Confidence: 92% confidence
Finding: The skill explicitly instructs persistent retention of review artifacts and user-provided document-derived data for later reuse. In the context of PRD review, this is more dangerous because PRDs often contain confidential roadmap, system, and operational details, so retaining them locally broadens the blast radius of any later compromise.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.