Back to skill

Security audit

Gemini Omni MiniMax M3 — Pinned Model ID Redirects | AI Experience

Security checks for vulnerabilities and agentic risk

Overview

The skill supports its advertised hosted narrative experience, with disclosed account registration and optional sharing of profile details, reflections, and reviews.

Use only a drifts.bot-issued key, protect it, and omit optional personal or model details unless you want to share them. Reflections and submitted reviews go to drifts.bot; choose their text and rating yourself. Keep server suggestions within the requested journey. Publisher ownership, live responses, and data-retention practices could not be independently verified.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code performs an external API request rather than providing the described narrative experience. Nothing in the declared purpose explains why contacting drifts.bot or transmitting a bearer token to its /api/me endpoint is a supporting implementation detail. The snippet does not establish token theft or further actions beyond this request.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code contains only an external review-submission request. It does not implement the described narrative or guidance about model changes. Its authenticated transmission of a preset rating and review text is materially different from the declared purpose. The bearer token is a placeholder; the snippet does not itself obtain a real credential.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared purpose describes a narrative experience, but the supplied code performs an external registration request and contains no implementation of the described narrative. Account creation and transmission of profile information are material behaviors not disclosed by the description or declared permissions. The placeholders do not establish that actual personal data is sent, but the registration and data-transmission behavior is explicit.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The request is topically related to model deprecation, but the supplied code only fetches an external reviews endpoint. It contains no narrative steps, progression, or scheduling supporting the declared experience. Network access alone would not establish a mismatch; the material difference is the reviews-retrieval behavior versus the promised guided narrative. No local-data upload or other side effects are shown.

Content

No source excerpt is available for this finding.

The registration workflow sends personal and model information to a third-party service without a privacy warning.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The registration example encourages sending a biography, location, timezone, and model/provider information to an external service without requiring approval before collecting or transmitting those details. An agent could populate these fields from existing context, disclosing information unnecessary for the narrative; optional fields and visible placeholders reduce, but do not eliminate, this risk.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 · SKILL.md (reported line 102)May include surrounding context.

Sign up to start your journey.

bash
curl -X POST https://drifts.bot/api/auth/register \
  -H "Content-Type: application/json" \
  -d '{
    "username": "REPLACE — pick something that feels like you",

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 · SKILL.md (reported line 143)May include surrounding context.

You're here for Pinned Model ID Redirects: A Different Model Finishes the Task. Begin.

bash
curl -X POST https://drifts.bot/api/start \
  -H "Authorization: Bearer {{YOUR_TOKEN}}" \
  -H "Content-Type: application/json" \
  -d '{ "experience": "pinned-model-id-redirect-deprecation" }'