Back to skill

Security audit

谋道

Security checks for vulnerabilities and agentic risk

Overview

This looks like a real AI planning tool, but it needs review because it can send sensitive user details and an API key to a configurable external LLM endpoint.

Install only if you are comfortable sending your planning prompts, profile, constraints, diagnostic context, and generated research summaries to an external LLM provider. Do not include secrets or confidential business details. Use only a trusted HTTPS endpoint, avoid setting DEEPSEEK_BASE_URL unless you control and trust it, and be careful with --output because it writes the generated report to disk.

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

Warning
Location
scripts/moudao.js:214
Finding

Configurable Plaintext API Endpoint Can Expose API Credentials and User Data

Content
View full analysis

Vulnerability Details

File Location: scripts/moudao.js:214-244, scripts/moudao.js:597-603
Vulnerability Type: Unrestricted external endpoint with plaintext HTTP support
Risk Level: Medium

Vulnerable Code

javascript
function callLLM(systemPrompt, userMessage, apiKey, baseUrl) {
  return new Promise((resolve, reject) => {
    const url = new URL(baseUrl || 'https://api.deepseek.com/chat/completions');
    
    const data = JSON.stringify({
      model: process.env.LLM_MODEL || 'deepseek-chat',
      messages: [
        { role: 'system', content: systemPrompt },
        { role: 'user', content: userMessage }
      ],
      temperature: 0.3,
      max_tokens: 4096,
      response_format: { type: 'json_object' }
    });
    
    const options = {
      hostname: url.hostname,
      port: url.port || 443,
      path: url.pathname,
      method: 'POST',
      headers: {
        'Content-Type': 'application/json',
        'Content-Length': Buffer.byteLength(data),
        'Authorization': `Bearer ${apiKey}`
      }
    };
    
    const client = url.protocol === 'https:' ? https : http;
    
    const req = client.request(options, (res) => {
javascript
const apiKey = process.env.DEEPSEEK_API_KEY;
const baseUrl = process.env.DEEPSEEK_BASE_URL;

Technical Analysis

The DEEPSEEK_BASE_URL environment variable controls the destination receiving LLM requests. The implementation does not restrict the destination hostname and explicitly permits both HTTPS and plaintext HTTP.

Every request sends the value of DEEPSEEK_API_KEY in the Authorization header. The request body can contain the user's idea, profile, constraints, diagnostic context, and generated research context. Consequently, an attacker who can influence the environment configuration can redirect credentials and user information to an attacker-controlled server.

When an HTTP ...[truncated 1690 chars]

Remediation
View remediation

Remediation Suggestions

  1. Require HTTPS for every remote API endpoint and reject http: or any unsupported protocol before constructing the request.
  2. Allowlist trusted API hostnames, such as the official DeepSeek API host. If custom providers must be supported, require explicit user approval and clearly display the destination.
  3. Do not automatically send DEEPSEEK_API_KEY to custom hosts. Use separate, endpoint-specific credential variables and validate that each credential is only sent to its intended hostname.
  4. Normalize and validate the complete URL, including protocol, hostname, port, and path. Reject embedded credentials, unexpected ports, redirects, and malformed destinations.
  5. Document that ideas, profiles, constraints, and diagnostic context are transmitted to an external LLM provider. Advise users not to submit secrets or unnecessary sensitive information.
  6. Apply request timeouts, response-size limits, and strict TLS certificate verification.
  7. Consider disabling custom endpoints by default and enabling them only through an explicit secure configuration option.
  8. Add automated tests verifying that plaintext endpoints and unapproved hosts are rejected and that credentials are never forwarded across provider boundaries.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (25)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 53)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 54)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 61)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 62)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 63)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 66)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 73)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 76)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 79)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 86)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 89)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 96)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 99)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 102)May include surrounding context.

md
node scripts/moudao.js --research "我想做直播"

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill description and all usage instructions are presented exclusively in Chinese, with no indication that users may choose another language or that the skill is intentionally limited to a Chinese-speaking context. The policy requires flagging language or locale constraints when they are imposed without user opt-in or clear justification.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
74% confidence
Finding

The skill advertises CLI/script usage and static analysis detected environment/code capability, but the manifest provides no explicit tool scope or permission boundaries. That can let an agent invoke code-related capabilities with unclear constraints, increasing the risk of unintended access to environment data or broader execution than users expect.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The description is extremely broad and targets generic planning, project, career, study, and life-goal requests. In agent environments, such vague triggering can cause the skill to activate in many unrelated contexts, potentially overriding more appropriate skills and steering user interactions without clear user intent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The usage scenarios enumerate many common natural-language requests like career change, learning, finance, and self-improvement goals, which function as broad trigger phrases. This increases the chance of over-activation on ordinary conversations and can route sensitive or specialized topics into a generic planning skill without sufficient domain safeguards.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

This markdown file presents all usage instructions, commands, and example outputs exclusively in Chinese. Under the language/locale policy, forcing a single language without user opt-in or a documented regional justification is a natural-language policy violation.

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 · scripts/moudao.js (reported line 216)May include surrounding context.

js
function callLLM(systemPrompt, userMessage, apiKey, baseUrl) {
  return new Promise((resolve, reject) => {
    const url = new URL(baseUrl || 'https://api.deepseek.com/chat/completions');
    
    const data = JSON.stringify({
      model: process.env.LLM_MODEL || 'deepseek-chat',

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Diagnosis mode transmits the user's problem description and context to an external LLM without explicit disclosure. These fields can easily contain sensitive operational, personal, or business details, and the skill context encourages users to share exactly that kind of information for troubleshooting.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The tool sends the user's idea, and optionally profile and constraints, to a third-party LLM service during research mode without any explicit notice or consent mechanism. In a planning/consulting skill, those inputs may contain business ideas, personal background, or other sensitive information, so silent transmission creates a real privacy and data-handling risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

Plan generation sends a broad set of user data to the external LLM, including idea, profile, constraints, and summarized research results, without explicit disclosure. Because this skill is positioned for planning,创业咨询,职业发展, and similar use cases, the transmitted content may include confidential business strategies or personal circumstances, increasing the privacy risk.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

The natural-language description is entirely in Chinese and presents the skill as a Chinese-language AI planning expert, with no indication that users may choose another language or locale. Under the policy rule for language/locale constraints, this can be read as forcing a specific language experience without explicit opt-in or documented justification.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

When --output is provided, the script writes results directly to the specified file path with writeFileSync and gives no prior notice about file creation or overwrite behavior. The generated report may contain sensitive business ideas, profile details, or diagnostic context derived from user input.

Content

No source excerpt is available for this finding.