Back to skill

Security audit

okx-cex-auth

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this skill suspicious because it reports credential-sharing guidance and actions that go beyond its described OKX authentication purpose; the scan was partial, so the evidence is incomplete.

Consider avoiding installation until the credential-sharing guidance and the mismatch about installing a trading CLI and logging out are clarified. Do not paste API keys, secret keys, or passphrases into chat. The scan covered the reported skill file but was marked partial because it could not resolve some referenced paths, so it does not establish that all related material was reviewed.

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
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (7)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied command performs logout rather than login, setup, status checking, or authentication recovery. Removing an authentication session is distinct from the declared removal of the okx-auth binary and can directly oppose the stated goal of establishing or restoring authenticated access.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

The description permits authentication setup and management of a named authentication binary, but the code installs a different, trading-oriented CLI package globally. Nothing supplied establishes that this package installation implements the declared authentication setup, so the installation target constitutes a scope mismatch.

Content

No source excerpt is available for this finding.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 194)May include surrounding context.

text

     **DO NOT include any other field from `auth status --json` in this reply.** Specifically:
     - `expiresAt` / `ttl` refer to the short-lived access token, not the OAuth session. The CLI auto-refreshes tokens transparently; surfacing these values will mislead users into thinking their login expires soon.
     - `profile` is an internal routing field with no user value.
     - Only `site` and `scopes` are user-relevant.
     - If asked about session longevity, say "Session stays active as long as you use the CLI periodically." Do not quote a number.

The skill requires a Chinese-language site menu regardless of the user's language or preference.

Medium
Category
Output Handling
Confidence
99% confidence
Finding

The instruction requires the agent to echo the site menu verbatim in Chinese, and the pre-login gate at lines 115–119 enforces that requirement for every login. No language opt-in or region-specific justification is provided; offering English login templates later does not remove this mandatory Chinese interaction.

Content

No source excerpt is available for this finding.

The API-key replacement workflow invites users to share account credentials in chat without a privacy warning.

Medium
Category
Output Handling
Confidence
99% confidence
Finding

The recovery workflow explicitly offers sharing the API key, secret key, and passphrase with the assistant, exposing a complete credential set to conversation history and potentially associated logging systems. In a financial trading context, compromised credentials could enable unauthorized account access or trading within the key's permissions; this is the same underlying issue as the SSD-3 finding.

Content

No source excerpt is available for this finding.

The recovery instructions invite users to disclose their API key, secret key, and passphrase directly to the assistant.

Medium
Category
Data Exfiltration
Confidence
99% confidence
Finding

Inviting users to send AK/SK/PP directly to the assistant unnecessarily moves sensitive authentication material into the chat environment. These credentials may authorize financial operations, making disclosure consequential even though the skill does not specify an attacker-controlled destination; this duplicates the credential-handling issue identified by SQP-2.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% 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.

Content

Scanner excerpt · SKILL.md (reported line 298)May include surrounding context.

bash
okx auth remove          # interactive confirmation
okx auth remove --force  # skip confirmation

Troubleshooting: "Failed to spawn okx-auth"