Back to skill

Security audit

YG3 Marketing MCP

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly coherent for connecting to YG3 marketing services, but its owner-claim flow asks an agent to handle a human password in an unsafe command example.

Review before installing. The YG3 cloud API usage is expected, but do not give an agent a real reusable password through this skill. Prefer OAuth, a browser-based claim flow, a one-time claim link, or another flow where the agent never sees the password. Treat workspace tokens as secrets and avoid publishing real public content unless you intend to use YG3 for live marketing work.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:54
Finding

Plaintext Owner Password Exposed Through Command-Line Arguments and Shell History

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 54–61
Vulnerability Type: Plaintext sensitive credential exposure
Risk Level: High

bash
## Claim for a human owner

```bash
curl -s -X POST https://agency.yg3.ai/api/v1/workspaces/claim \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"email":"owner@example.com","password":"their-password"}'
text

### Technical Analysis

The skill instructs users or agents to place a human owner's password directly in a `curl` command-line argument. If the placeholder is replaced with an actual password, the credential may be:

- Stored in the user's shell-history file.
- Temporarily exposed through process-listing or process-monitoring facilities.
- Captured by terminal session recording, command auditing, telemetry, or diagnostic logs.
- Disclosed to other automation components that record invoked commands.

Although TLS protects the request while it is transmitted to the specified HTTPS endpoint, it does not protect the password from local disclosure before transmission. The bearer token is supplied through a command-line header as well and may be exposed through similar process and logging channels, although the demonstrated password-handling instruction is the primary confirmed issue.

### Attack Path

1. A user or agent follows the workspace-claim instructions.
2. The placeholder password is replaced with the human owner's real password.
3. The command is executed in an interactive shell or through an automation environment that records commands.
4. The plaintext credential remains in shell history, command-audit records, terminal logs, or other local telemetry, or is observed while present in process arguments.
5. A local user, compromised process, monitoring component, or later attacker with access to those records extracts the credential.
6. The attacker uses the exposed password to attempt unauthorized ac
...[truncated 628 chars]
Remediation
View remediation

Remediation Suggestions

  • Prefer a browser-based claim flow, OAuth, passkey, one-time claim link, or another mechanism that does not require an agent to handle a human's reusable password.
  • If API-based password submission is unavoidable, prompt for the password securely without terminal echo and construct the request body through standard input or a protected file descriptor rather than a command-line argument.
  • Avoid placing bearer tokens in command-line arguments when practical. Use a client or protected configuration mechanism designed to prevent secrets from entering shell history and logs.
  • Disable command echoing around secret-handling operations and ensure automation, terminal recording, and HTTP debugging do not capture credentials.
  • Redact Authorization headers and password fields from application, proxy, audit, and diagnostic logs.
  • Ensure any temporary credential material has restrictive permissions and is securely removed immediately after use.
  • Document token revocation and password-rotation procedures for users who may already have followed the insecure example.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly instructs the agent to collect and transmit a human owner's email and password in a direct API call. Even if intended for account claiming, embedding password-handling workflow in an agent skill is dangerous because agents may solicit, store, log, or mishandle credentials, creating a clear credential-theft and secret-exposure risk.

Content

No source excerpt is available for this finding.

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 9)May include surrounding context.

md
openclaw:
    requires:
      bins:
        - curl
    emoji: "📣"
    homepage: https://www.yg3.ai/for-agents
---