ClawHub
Back to skill

Security audit

Qwen

Security checks for vulnerabilities and agentic risk

Overview

This skill is a transparent Qwen setup and debugging guide that discloses its cloud API use and only saves local notes after user approval.

Install this only if you are comfortable with Qwen prompts, and any requested images, being sent to Alibaba Cloud Model Studio when you choose a hosted route. Keep sensitive workloads on a local endpoint, keep DASHSCOPE_API_KEY in environment variables, and approve ~/qwen/ notes only if you want persistent routing memory.

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
  • 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
Findings (5)

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
- stores lightweight local notes only after user approval

This skill NEVER:
- invent live model availability without checking
- persist secrets in `~/qwen/`
- execute destructive downstream automation without validated output
- pretend one backend's tool-calling behavior applies everywhere
Confidence
75% 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.

External Transmission

Medium
Category
Data Exfiltration
Content
## Minimal Hosted Chat Completion

```bash
curl -s "${QWEN_BASE_URL}/chat/completions" \
  -H "Authorization: Bearer $DASHSCOPE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
## Minimal Hosted Chat Completion

```bash
curl -s "${QWEN_BASE_URL}/chat/completions" \
  -H "Authorization: Bearer $DASHSCOPE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
## Minimal Local Chat Check

```bash
curl -s "${LOCAL_BASE_URL}/chat/completions" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MODEL_ID",
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The activation criteria are broad enough to trigger on several related product or backend names, which can cause the skill to engage in contexts the user did not clearly intend. Over-broad activation increases the chance of unsolicited persistence prompts, routing advice, or file-write workflows appearing in unrelated conversations, which can create privacy and integrity risks through accidental scope expansion.

Static analysis

No suspicious patterns detected.