AIスライド作成ワークフロー

• Outcome: Review
• Latest audit: May 29, 2026 at 9:22 PM UTC
• Version:
  1.1.0

Overview

This slide-making skill is mostly legitimate, but it asks the agent to bypass permissions and can send or publicly share presentation files through hard-coded external accounts.

Review before installing. Remove the permission-bypass flag, replace hard-coded Google credentials and folder IDs with your own private configuration, keep outputs local by default, stop the temporary server after export, and only upload or publicly share decks after explicit confirmation that the content can be exposed.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Context-Inappropriate Capability

Medium

• Confidence: 94% confidence
• Finding: The skill includes Google Drive upload plus `share ... --to anyone --role reader`, which creates a public link to generated artifacts. For a slide-creation workflow, broad public sharing is not necessary by default and can expose confidential client content, branding, or embedded sensitive data to unintended parties.

Context-Inappropriate Capability

High

• Confidence: 98% confidence
• Finding: The workflow invokes Claude Code with `--dangerously-skip-permissions`, disabling an important safety boundary and allowing the downstream tool to act with excessive privileges. Because the prompt body contains user-provided materials and the tool can generate or execute project actions, this raises the risk of unintended file access, command execution, or data exfiltration beyond what slide generation requires.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The workflow includes Telegram delivery and optional Google Drive upload without clear, user-facing warnings that materials may be transmitted to third-party services. In a slide workflow, files may contain client names, internal strategy, screenshots, or proprietary visuals, so silent or assumed disclosure meaningfully increases confidentiality risk.

Ssd 3

Medium

• Confidence: 91% confidence
• Finding: The skill explicitly instructs users to send outlines, design prompts, character images, and client/context details to external AI services. This normalizes disclosure of potentially sensitive business information to third parties without clear minimization or consent safeguards, which is risky in a client-deliverable workflow where materials may be confidential.

Ssd 3

Medium

• Confidence: 93% confidence
• Finding: Sending generated deliverables to Telegram and optionally uploading them to Google Drive extends the exposure surface for user work product. In this context, the workflow handles presentation files that may contain confidential client strategy, branding, or personal data, so routine external transmission without strict controls is more dangerous than in a purely public-content workflow.

VirusTotal

65/65 vendors flagged this skill as clean.