ClawHub
Back to skill

Security audit

Darktable Photo Workflow

Security checks across malware telemetry and agentic risk

Overview

This is a self-contained Darktable photo-analysis skill with limited, disclosed optional browsing for documentation and art-history verification.

Install is reasonable if you want a structured Darktable photo-editing assistant. Be aware that although the package itself is just local Markdown/YAML, the instructions may ask the agent to browse official documentation or art-history sources when current or exact facts are needed; disable browsing if you require fully offline use.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
93% confidence
Finding
The manifest explicitly claims the skill works offline with zero dependencies, but the body permits live web verification in some cases. This creates a trust-boundary mismatch: users or orchestrators may enable the skill expecting no network access, while the skill can still trigger external requests and disclose prompts, filenames, metadata, or user intent to third-party resources.

Description-Behavior Mismatch

Medium
Confidence
93% confidence
Finding
The manifest explicitly claims the skill works offline with zero dependencies, but the body permits live web verification in some cases. This creates a trust-boundary mismatch: users or orchestrators may enable the skill expecting no network access, while the skill can still trigger external requests and disclose prompts, filenames, metadata, or user intent to third-party resources.

Context-Inappropriate Capability

Medium
Confidence
90% confidence
Finding
The skill enables web-assisted live research despite being framed as an offline editing workflow. Unnecessary network access increases attack surface and can leak user-provided images, metadata, prompts, or sensitive project context to external sites, especially if the agent follows these instructions automatically during analysis.

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
Instructing the agent to use web tools to verify manuals and release notes contradicts the stated offline/self-contained behavior and can cause external data transmission. This is dangerous because version checks may send user context or file-derived details off-device, and deployment systems may grant network permissions based on the manifest's offline claims.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.