Back to skill

Security audit

OpenRouter

Security checks for vulnerabilities and agentic risk

Overview

This skill needs review because it treats potentially paid chat generation as a safe read and directs immediate execution of an unverified installer during setup.

Before use, verify and install the CLI yourself and review its automatic agent-skill installation behavior. Explicitly authorize the model, prompt content, and spending limits for chat generation; do not rely on this skill's safe-read label. Prompts pass through OOMOL to OpenRouter and model providers. The bundled instructions were fully reviewed, but the external installer and hosted implementation were not verified. No malicious behavior was 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (5)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied command runs an installer from a non-OpenRouter resource. Nothing in the code establishes that this installation supports the declared OpenRouter functionality. Executing this remote installer is therefore an undeclared capability, not an identifiable supporting implementation detail. The downloaded script's contents and further effects are not visible in the supplied chunk.

Content

No source excerpt is available for this finding.

External Script Fetching

High
Category
Supply Chain
Confidence
95% confidence
Finding

The setup command executes a mutable remote script immediately without an independently verified checksum, signature, or review; the PowerShell alternative has the same weakness. A compromised distribution endpoint could execute arbitrary code with the invoking user's privileges, although the missing-CLI condition limits exposure and no malicious intent is evident.

Content

Scanner excerpt · SKILL.md (reported line 67)May include surrounding context.

The chat-completion action is classified as a safe read despite creating a billable generation.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The untagged create_chat_completion action inherits the explicit rule that untagged actions are safe reads, although it sends prompts externally and can consume paid credits. This bypasses the skill's confirmation requirement for consequential operations; an explicit user request may authorize a particular generation, but the blanket classification does not establish that authorization.

Content

No source excerpt is available for this finding.

The safety guidance treats chat generation as a safe read without warning about prompt disclosure or usage charges.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The safety guidance conflates untagged operations with harmless reads, leaving chat completion outside the safeguards applied to create_message. Consequently, an agent could transmit sensitive prompt content or incur charges without checking appropriate authorization; this overlaps with SDI-4 rather than representing an independent exploit.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 88)May include surrounding context.

md
- **`OOMOL_INSUFFICIENT_CREDIT`** — OOMOL billing stop. Recharge at `https://console.oomol.com/billing/token-recharge` before retrying.

- **HTTP 402 without `OOMOL_INSUFFICIENT_CREDIT`** — the third-party account itself has no active plan or credit. Report the provider error message as is; do not ask the user to recharge OOMOL.

## Resources