ClawHub
Back to skill

Security audit

Ollama Model Tuner

Security checks for vulnerabilities and agentic risk

Overview

This skill is a small, purpose-aligned Ollama tuning helper that reads a user-provided dataset and sends sample prompts to a local Ollama model, with no evidence of hidden persistence, exfiltration, or destructive behavior.

Before installing, confirm you have Ollama configured as intended and do not use datasets containing secrets, personal data, or proprietary text unless you are comfortable processing that data through your local Ollama setup. Expect some rough packaging: the documented command wrapper and prompts/system.md reference are not present in the artifact.

A.I.G

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

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

External Model or Provider Selection

High
Category
Excessive Agency
Content
## 🚀 Quick Start
```
!ollama-model-tuner --model llama3 --dataset ./data.json --task classification
```

## Files
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.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The script sends dataset-derived prompts directly to an external/local Ollama model service via `ollama.chat` without any notice, consent flow, or data-sensitivity checks. If the dataset contains proprietary, personal, or otherwise sensitive prompt content, this can cause unintended disclosure to another service boundary and create privacy/compliance issues.

Static analysis

No suspicious patterns detected.