ClawHub
Back to skill

Security audit

RamaLama CLI

Security checks for vulnerabilities and agentic risk

Overview

This skill is a straightforward RamaLama CLI guide with expected local model, container, RAG, and serving commands, though users should be mindful when starting a service endpoint.

Install only if you intend to use RamaLama with Docker or Podman. Be careful when using RAG paths or prompts that contain private data, and when running `ramalama serve -d`, verify the host/port exposure and stop the service when finished.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Confidence
93% confidence
Finding
The phrase "Use when an alternative AI agent is better suited to a task" is a very broad trigger condition that lacks clear boundaries for when this skill should or should not be used. Although examples are provided, there is no explicit trigger list or negative examples to prevent overlap with many ordinary agent tasks.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The skill recommends `ramalama serve` and notes that it exposes an OpenAI-compatible endpoint, but it does not prominently warn that serving opens a local API surface that other local processes, containers, or potentially network peers may reach depending on bind settings/runtime defaults. In a skill specifically designed to run AI agents and process potentially sensitive prompts, omission of that warning can lead to unintended data exposure or unauthorized use of the model service.

External Transmission

Medium
Category
Data Exfiltration
Content
```bash
ramalama serve -d granite3.3:2b
curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"granite3.3:2b","messages":[{"role":"user","content":"Hello"}]}'
```
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Static analysis

No suspicious patterns detected.