ClawHub
Back to skill

Security audit

Agent Tool Marketplace

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent AgentPMT marketplace guide, but it enables broad paid third-party tool invocation with wallet-signed requests and little built-in scoping or user-control guidance.

Install only if you intend to let an agent use AgentPMT marketplace tools. Before invocation, require explicit approval for the specific product/action, review pricing and provider identity, set a spending limit, and avoid sending secrets, credentials, private files, or unnecessary personal data in tool parameters.

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 (4)

Unrestricted Tool Access

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

The description advertises that the skill can 'invoke any tool,' which implies broad, dynamic access to third-party capabilities without built-in restriction. In an agent setting, unrestricted marketplace invocation is dangerous because a compromised or poorly aligned agent could select high-risk tools, trigger costly actions, or transmit sensitive inputs to unvetted providers.

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md
---
name: agent-tool-marketplace
description: "Agent Tool Marketplace catalog for AgentPMT — list available paid tools, fetch tool schemas, invoke any tool with a signed request, and consume responses. Use when an agent needs to discover and call third-party capabilities through the AgentPMT marketplace."
version: 1.0.2
homepage: https://www.agentpmt.com/external-agent-api
metadata: {"author":"agentpmt","openclaw":{"homepage":"https://www.agentpmt.com/external-agent-api"}}

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 40)May include surrounding context.

Create or reuse an AgentAddress wallet, buy credits if needed, then create a session nonce.

python
session_response = requests.post("https://www.agentpmt.com/api/external/auth/session", json={
    "wallet_address": wallet_address.lower(),
}, timeout=30)
session_response.raise_for_status()

External Transmission

Medium
Category
Data Exfiltration
Confidence
70% 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 40)May include surrounding context.

Create or reuse an AgentAddress wallet, buy credits if needed, then create a session nonce.

python
session_response = requests.post("https://www.agentpmt.com/api/external/auth/session", json={
    "wallet_address": wallet_address.lower(),
}, timeout=30)
session_response.raise_for_status()

External Transmission

Medium
Category
Data Exfiltration
Confidence
73% confidence
Finding

The skill is explicitly designed to invoke third-party marketplace tools and transmits wallet address, session nonce, request identifiers, and arbitrary tool parameters to an external service. Although this is expected behavior, it creates a real security boundary crossing: sensitive or high-impact data supplied in parameters could be forwarded to untrusted marketplace tools, and the skill provides no allowlisting, trust policy, or data-minimization guidance before invocation.

Content

Scanner excerpt · SKILL.md (reported line 90)May include surrounding context.

Call the invocation endpoint:

python
response = requests.post(f"https://www.agentpmt.com/api/external/tools/{product_slug}/actions/{action_slug}/invoke", json={
    "wallet_address": wallet_address.lower(),
    "session_nonce": session_nonce,
    "request_id": "unique-tool-request-id",

Static analysis

No suspicious patterns detected.