MasterSwarm AI Document Analysis

• Outcome: Pass
• Latest audit: May 28, 2026 · 8:41 AM UTC
• Version: 1.1.2

Overview

This skill is a disclosed cloud API connector, but users should only send documents they are comfortable sharing with MasterSwarm/NeuroDoc.

Install only if you are comfortable sending selected text or documents to MasterSwarm/NeuroDoc. Redact secrets and unnecessary personal data, be cautious with medical/legal/financial documents, verify the provider's privacy and retention claims independently, and confirm before multi-engine runs because they can send the same content multiple times and consume paid credits.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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 (8)

Vague Triggers

Medium

• Confidence: 92% confidence
• Finding: The skill advertises analysis of "any document" with no scope limits or safety boundaries, which can cause over-broad activation and encourage sending highly sensitive material to a third-party API. In this context, the ambiguity is more dangerous because the skill explicitly supports receipts, contracts, lab results, legal, business, and crypto use cases, all of which commonly contain confidential data.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: Telling users to send "any text or question" creates an ambiguous activation condition that can capture secrets, regulated data, or unrelated private content without meaningful boundaries. Because this skill is an API connector that forwards inputs to an external service, the broad phrasing materially increases the chance of unsafe data exfiltration.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The manifest requires a MasterSwarm API key and advertises cloud-based document analysis, but it does not disclose credential handling practices or clearly warn that user documents and prompts may be transmitted to a third-party service. Because the skill targets sensitive content such as receipts, contracts, medical, legal, and crypto-related documents, the missing disclosure materially increases the risk of users exposing confidential data without informed consent.

External Transmission

Medium

• Category: Data Exfiltration

Content

### Single Engine Analysis

```bash
curl -s --max-time 120 -X POST https://api.neurodoc.app/aetherlang/execute \
  -H "Content-Type: application/json" \
  -H "X-Aether-Key: MASTERSWARM_API_KEY" \
  -d '{

• Confidence: 96% confidence
• Finding: The direct reference to api.neurodoc.app at the submission step confirms outbound transmission at the exact point where user content is posted. In context, the danger is elevated because the skill encourages uploading documents that often contain PII, financial, medical, or contractual data, and users may interpret marketing claims like "zero retention" as sufficient assurance without independent verification.

External Transmission

Medium

• Category: Data Exfiltration

Content

```bash
# Example: Run apex + consult + research on the same query
for ENGINE_ID in apex consult research; do
  curl -s --max-time 120 -X POST https://api.neurodoc.app/aetherlang/execute \
    -H "Content-Type: application/json" \
    -H "X-Aether-Key: MASTERSWARM_API_KEY" \
    -d "{

• Confidence: 96% confidence
• Finding: The multi-engine loop again references the external host at the point of repeated transmission, reinforcing that sensitive user content may be sent externally several times in one workflow. The repeated-send design makes the context more dangerous than a generic API connector because the skill explicitly frames this as a comprehensive analysis feature for potentially confidential documents.

External Transmission

Medium

• Category: Data Exfiltration

Content

homepage: https://masterswarm.net
  skill_type: api_connector
  external_endpoints:
    - https://api.neurodoc.app/aetherlang/execute
  operator_note: "api.neurodoc.app and masterswarm.net are the same operator (NeuroDoc Pro, Hetzner DE)"
license: MIT
---

• Confidence: 94% confidence
• Finding: The metadata declares an external endpoint, confirming that the skill is designed to send data off-platform to a remote operator. While not hidden, this is still a true security concern because the skill markets processing of broad, potentially sensitive document types and asks users to trust privacy assertions made by the same operator.

External Transmission

Medium

• Category: Data Exfiltration

Content

### Single Engine Analysis

```bash
curl -s --max-time 120 -X POST https://api.neurodoc.app/aetherlang/execute \
  -H "Content-Type: application/json" \
  -H "X-Aether-Key: MASTERSWARM_API_KEY" \
  -d '{

• Confidence: 96% confidence
• Finding: The direct reference to api.neurodoc.app at the submission step confirms outbound transmission at the exact point where user content is posted. In context, the danger is elevated because the skill encourages uploading documents that often contain PII, financial, medical, or contractual data, and users may interpret marketing claims like "zero retention" as sufficient assurance without independent verification.

External Transmission

Medium

• Category: Data Exfiltration

Content

```bash
# Example: Run apex + consult + research on the same query
for ENGINE_ID in apex consult research; do
  curl -s --max-time 120 -X POST https://api.neurodoc.app/aetherlang/execute \
    -H "Content-Type: application/json" \
    -H "X-Aether-Key: MASTERSWARM_API_KEY" \
    -d "{

• Confidence: 96% confidence
• Finding: The multi-engine loop again references the external host at the point of repeated transmission, reinforcing that sensitive user content may be sent externally several times in one workflow. The repeated-send design makes the context more dangerous than a generic API connector because the skill explicitly frames this as a comprehensive analysis feature for potentially confidential documents.

VirusTotal

64/64 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.