Back to skill

Security audit

minimax-delegation

Security checks for vulnerabilities and agentic risk

Overview

The verdict is suspicious because scanners flag a risk that files may be sent to MiniMax without a clear privacy warning or sensitivity check; the scan was partial, so evidence is incomplete.

Before installing, consider whether you are comfortable with selected file contents being sent to MiniMax and with commands choosing a specific model. The scan is partial: it inspected two Markdown components but could not resolve several references, so some referenced behavior was not established.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (8)

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
90% confidence
Finding

Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.

Content

Scanner excerpt · SKILL.md (reported line 104)May include surrounding context.

bash
# Basic command
mmx text chat --model MiniMax-M3.1-Flash --message "Analyze this code"

# Specific model
mmx text chat --model MiniMax-M3 --message "Summarize this design"

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
90% confidence
Finding

Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.

Content

Scanner excerpt · SKILL.md (reported line 107)May include surrounding context.

md
mmx text chat --model MiniMax-M3.1-Flash --message "Analyze this code"

# Specific model
mmx text chat --model MiniMax-M3 --message "Summarize this design"

# JSON output
mmx text chat --message "Extract the function names" --output json

File delegation lacks a warning about transmitting potentially sensitive source content to MiniMax.

Medium
Category
Output Handling
Confidence
86% confidence
Finding

The workflow inlines requested local files into prompts transmitted to MiniMax without documenting a sensitivity check, redaction step, or confirmation of external disclosure. Although external delegation is the skill's intended function, broad file-analysis requests can inadvertently disclose credentials, proprietary code, or personal information; the referenced shared delegation core might provide protections, but those are not established here.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · modules/minimax-specifics.md (reported line 63)May include surrounding context.

md
| Region | OpenAI-compatible | Anthropic-compatible | Documentation |
|--------|-------------------|----------------------|---------------|
| Global | `https://api.minimax.io/v1` | `https://api.minimax.io/anthropic` | `https://platform.minimax.io/docs` |
| China | `https://api.minimaxi.com/v1` | `https://api.minimaxi.com/anthropic` | `https://platform.minimaxi.com/docs` |

Keys are issued per region and do not work across regions. API-key login

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · modules/minimax-specifics.md (reported line 63)May include surrounding context.

md
| Region | OpenAI-compatible | Anthropic-compatible | Documentation |
|--------|-------------------|----------------------|---------------|
| Global | `https://api.minimax.io/v1` | `https://api.minimax.io/anthropic` | `https://platform.minimax.io/docs` |
| China | `https://api.minimaxi.com/v1` | `https://api.minimaxi.com/anthropic` | `https://platform.minimaxi.com/docs` |

Keys are issued per region and do not work across regions. API-key login

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · modules/minimax-specifics.md (reported line 64)May include surrounding context.

md
| Region | OpenAI-compatible | Anthropic-compatible | Documentation |
|--------|-------------------|----------------------|---------------|
| Global | `https://api.minimax.io/v1` | `https://api.minimax.io/anthropic` | `https://platform.minimax.io/docs` |
| China | `https://api.minimaxi.com/v1` | `https://api.minimaxi.com/anthropic` | `https://platform.minimaxi.com/docs` |

Keys are issued per region and do not work across regions. API-key login
probes both regions and saves the one that accepts the key, so the base

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · modules/minimax-specifics.md (reported line 64)May include surrounding context.

md
| Region | OpenAI-compatible | Anthropic-compatible | Documentation |
|--------|-------------------|----------------------|---------------|
| Global | `https://api.minimax.io/v1` | `https://api.minimax.io/anthropic` | `https://platform.minimax.io/docs` |
| China | `https://api.minimaxi.com/v1` | `https://api.minimaxi.com/anthropic` | `https://platform.minimaxi.com/docs` |

Keys are issued per region and do not work across regions. API-key login
probes both regions and saves the one that accepts the key, so the base

The context-inclusion workflow instructs sending file contents to MiniMax without a privacy warning.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The context-inclusion instructions direct users to transmit file contents and split larger corpora across requests without documenting a privacy warning or content-screening requirement. This creates an accidental disclosure risk for sensitive files; the 96 KiB limit controls argument size, not confidentiality.

Content

No source excerpt is available for this finding.