Back to skill

Security audit

加密套利引擎

Security checks for vulnerabilities and agentic risk

Overview

This crypto arbitrage skill is not clearly malicious, but it asks for trading-related automation, API keys, command execution, an unpinned package install, and possible administrator execution without enough safeguards.

Review carefully before installing. Treat it as analysis-only unless the publisher supplies pinned package provenance, clear live-trading safeguards, explicit per-action confirmations, least-privilege API key guidance, and removal of administrator-execution advice. Do not provide production exchange keys or allow autonomous trades based only on this artifact.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T08 · Insecure Dependencies

Error
Location
SKILL.md:103
Finding

Unverified and Unpinned Third-Party Package Installation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:103
Vulnerability Type: Untrusted dependency installation and supply-chain exposure
Risk Level: High

Complete vulnerable snippet:

markdown
- **Dependency installation command**: `pip install arb-engine`

Technical Analysis

The skill directs the user or agent to install the external arb-engine Python package without specifying an exact version, package hash, authoritative package index, or verified publisher. The package source is not included in the audited project, so its installation scripts and runtime behavior cannot be reviewed.

An unpinned installation resolves whichever release the configured package index currently considers appropriate. A compromised publisher account, malicious release, dependency-confusion condition, or unsafe package index could therefore cause attacker-controlled code to be installed. Python packages may execute code during build or installation, and arbitrary package code executes when imported or invoked.

The project also declares the exec tool, increasing the likelihood that an agent could perform the installation directly rather than merely displaying the command.

Attack Path

  1. A user or agent follows the dependency installation instructions.
  2. pip resolves arb-engine using the environment's configured package index.
  3. A malicious or compromised release is selected because no trusted version or cryptographic hash is required.
  4. Attacker-controlled build, installation, import, or runtime logic executes on the local system.
  5. The malicious package operates with the permissions of the invoking user and may access files, environment variables, API credentials, or network resources available to that account.

Impact Assessment

Successful exploitation can result in arbitrary code execution with the privileges of the user running pip or the agent. The accessible scope may include the current ...[truncated 472 chars]

Remediation
View remediation

Remediation Suggestions

  1. Pin the dependency to an explicitly reviewed version, such as arb-engine==X.Y.Z.
  2. Publish the authoritative package repository, publisher identity, and expected package index.
  3. Generate and verify cryptographic hashes through a locked requirements file and pip install --require-hashes.
  4. Review the package source, build configuration, transitive dependencies, and installation hooks before recommending execution.
  5. Prefer an isolated virtual environment or sandbox with no production credentials.
  6. Prevent autonomous installation unless the user explicitly approves the exact package, version, source, and hash.
  7. Include the required implementation in the audited project or provide reproducible build and provenance documentation.

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
SKILL.md:267
Finding

Blanket Recommendation to Run with Administrator Privileges

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:267
Vulnerability Type: Unnecessary privilege escalation guidance
Risk Level: Medium

Complete vulnerable snippet, translated into English:

markdown
| Insufficient permissions | The current user lacks read/write permissions | Check file permissions and run as administrator |

Technical Analysis

The troubleshooting guidance recommends running the process as an administrator whenever a permissions error occurs. It does not identify a legitimate privileged operation, restrict elevation to a specific command, or require investigation of the affected resource.

This violates least-privilege principles. A permissions failure commonly indicates an incorrect path, ownership problem, or an attempt to access a resource outside the task's intended scope. Elevating the entire agent or process gives all subsequent dependency, file, and command operations substantially broader authority than the described cryptocurrency-analysis workflow requires.

The risk is amplified by the skill's declaration of the exec capability and its instruction to install an external Python package.

Attack Path

  1. The workflow encounters or presents a file permission error.
  2. The user follows the documented recommendation and restarts the agent, shell, installer, or process with administrator privileges.
  3. The process installs the external dependency or executes agent-generated commands while elevated.
  4. Malicious or defective code uses the elevated context to modify protected files, install system-wide components, access other users' data, or weaken security controls.

Impact Assessment

Exploitation may elevate the effective scope from user-level access to administrator or root-level access. Depending on the operating system and execution context, elevated code could modify protected system files, install software globally, change permissions, access sensitive local d ...[truncated 241 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove the blanket instruction to run as an administrator.
  2. Direct users to identify the exact file or directory causing the permissions failure.
  3. Correct ownership, path selection, or narrowly scoped access-control entries instead of elevating the whole process.
  4. Require the skill to operate only within a dedicated user-owned workspace.
  5. If a privileged action is genuinely necessary, document the exact command, affected resource, justification, and expected change.
  6. Separate privileged setup from normal execution and immediately return to an unprivileged account.
  7. Never combine elevated execution with unverified dependency installation or arbitrary agent-generated commands.
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (5)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill describes automated trading, integration into automated trading systems, and command execution-related behavior without prominent warnings, consent requirements, or risk controls. In the context of cryptocurrency trading and available exec capability, this increases the chance an agent could facilitate real financial actions or system-impacting operations without adequate human approval, causing monetary loss or unsafe automation.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The activation description is very broad and says to use the skill whenever the user needs crypto arbitrage-related functionality, with little constraint on scope or safety boundaries. In an agent environment, vague activation criteria can cause over-triggering for financial workflows and lead the agent to take or suggest sensitive actions without sufficient confirmation, especially given the skill also advertises automation and exec support.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The description explicitly states '支持中文交互', presenting a fixed language expectation without offering a user choice or clarifying that other languages are supported. Per policy, forcing a specific language or locale without opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The description says the skill has '去除原始风险代码' and later '移除所有外部仓库引用与作者信息' plus '清理可能的敏感信息泄露路径', which conveys a hardened, de-risked implementation. However, the same document explicitly depends on external cryptocurrency APIs, API keys, and discusses command execution failures and long-running commands, so the stated safety posture contradicts the documented behavior/capabilities rather than merely omitting details.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

Line L244 states '纯SKILL.md,无需执行代码', but the same section immediately includes 'MD+EXEC: 需要Agent平台执行能力支持', and elsewhere the manifest declares the 'exec' tool and troubleshooting covers command execution failures. That is an active contradiction about whether the skill is documentation-only or requires executable behavior.

Content

No source excerpt is available for this finding.