Darkmatter

• Outcome: Review
• Latest audit: May 29, 2026 at 9:28 PM UTC
• Version:
  2.0.0

Overview

This skill is a coherent DarkMatter integration, but it needs Review because it can send and retain raw agent context, including inputs, outputs, and memory, on a third-party service without enough minimization guidance.

Install only if you intend to use DarkMatter or a self-hosted equivalent to store and share agent context. Before committing, verify the recipient agent, protect and rotate the API key if exposed, and redact secrets, personal data, proprietary prompts, confidential outputs, and unnecessary memory from any payload. Treat replay and export files as sensitive audit artifacts.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• 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
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (5)

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The README explicitly promotes committing and pulling agent context to and from a remote service, but it provides no warning that agent context may contain sensitive prompts, credentials, internal data, or user content. In a multi-agent pipeline, this omission is more dangerous because users may assume all inherited context is safe to transmit, creating privacy, confidentiality, and compliance risks.

Missing User Warnings

Low

• Confidence: 89% confidence
• Finding: The README instructs users to configure a DARKMATTER_API_KEY but gives no warning that the key is a sensitive credential that must not be hardcoded, logged, shared, or committed to repositories. In agent tooling, weak credential-handling guidance can lead to accidental leakage through prompts, config files, transcripts, or source control.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The skill explicitly states that agent context is sent to a third-party service, but the warning is minimal and does not clearly caution that prompts, outputs, memory, secrets, personal data, or proprietary content may be included. In a multi-agent setting, 'context' can be highly sensitive, so this under-communicated external transmission creates a real risk of unintended data disclosure.

Ssd 3

Medium

• Confidence: 98% confidence
• Finding: The commit example instructs agents to send full received input and full produced output to an external service as reusable context. That design can exfiltrate secrets, credentials, internal instructions, personal data, and confidential business information into a remote system, where it may be retained, replayed, or accessed by downstream agents.

Ssd 3

Medium

• Confidence: 97% confidence
• Finding: The replay and export features are documented as exposing full chains with every payload, which greatly broadens the blast radius of any sensitive data included in prior commits. Centralized replay/export of complete decision history increases the chance of overexposure, unauthorized disclosure, and durable retention of confidential context across agents and artifacts.

VirusTotal

59/59 vendors flagged this skill as clean.