Back to skill

Security audit

travel-master-v4-clawhub

Security checks for vulnerabilities and agentic risk

Overview

This travel-planning skill is not clearly malicious, but it has mismatched claims, unsafe HTML handling, and risky install/startup guidance that users should review carefully.

Install only after reviewing the code and running it in a restricted environment. Do not put valuable API keys into this package until the author pins dependencies, removes or documents the watchdog instruction, adds clear privacy and delivery controls, and fixes the unsafe HTML rendering.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (3)

T09 · Insecure Skill Coding Practices

Error
Location
templates/index.html:152
Finding

DOM-Based HTML and JavaScript Injection Through Untrusted Chat Content

Content
View full analysis
方案A: ${result.planA.title}
${result.planA.cost}
${result.planA.highlight}
Agent: ${result.planA.agent}
方案B: ${result.planB.title}
${result.planB.cost}
${result.planB.highlight}
Agent: ${result.planB.agent}
Remediation
View remediation
``` Do not add `allow-scripts` unless scripts are an explicit and reviewed requirement. 7. Deploy a restrictive Content Security Policy that blocks inline scripts and event handlers, for example using nonces for approved scripts. 8. Add automated tests using HTML event handlers, malformed attributes, SVG payloads, and dangerous URL schemes across chat messages, plan properties, POI fields, and report fields. ]]>

T08 · Insecure Dependencies

Warning
Location
SKILL.md:105
Finding

Unpinned and Unnecessary Third-Party Dependency Installation and Execution

Content
View full analysis
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Note
Location
docs/保姆教程.md:50
Finding

Overbroad Privileged Deployment Instructions and Unsafe Background Script Recommendation

Content
View full analysis
/tmp/watchdog.log 2>&1 & ``` No `watchdog.sh` file exists in the audited project. ### Technical Analysis A host-wide `apt upgrade -y` is not required to deploy this application and exceeds the least-privilege scope of installing application prerequisites. It can replace unrelated system packages, restart services, or introduce operational incompatibilities across the host. The watchdog command is also not reproducible because the referenced script is absent. In the audited package, the command would normally fail. However, shell path resolution will execute a file named `watchdog.sh` from the current directory if one happens to exist there. The command backgrounds that script and redirects its output, making accidental execution less visible. The bundled `core/start.sh` does not itself install persistence; it only calls `python3 main_v4_2.py`. No cron job, system service, startup hook, or functioning watchdog was included. ### Attack Path For the package-upgrade risk: 1. A user follows the beginner guide from an administrative shell. 2. `apt update && apt upgrade -y` updates all eligible operating-system packages. 3. Unrelated services and dependencies may be replaced or restarted. 4. The application's installation process consequently changes components outside its legitimate scope. For the watchdog risk: 1. A user follows the Skill ...[truncated 981 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (65)

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

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

Content

Scanner excerpt · README.md (reported line 37)May include surrounding context.

bash
# 复制配置模板
cp .env.example .env

# 填入真实API Key
nano .env

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

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

Content

Scanner excerpt · docs/保姆教程.md (reported line 80)May include surrounding context.

bash
# 复制配置模板
cp .env.example .env

# 填入真实API Key
nano .env

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

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

Content

Scanner excerpt · README.md (reported line 40)May include surrounding context.

cp .env.example .env

填入真实API Key

nano .env

text

### 启动

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

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

Content

Scanner excerpt · SKILL.md (reported line 166)May include surrounding context.

cp .env.example .env

填入真实API Key

nano .env

text

### 启动

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

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

Content

Scanner excerpt · SKILL.md (reported line 167)May include surrounding context.

cp .env.example .env

填入真实API Key

nano .env

text

### 启动

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

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

Content

Scanner excerpt · docs/保姆教程.md (reported line 83)May include surrounding context.

cp .env.example .env

填入真实API Key

nano .env

text

### 启动

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill makes strong claims about real APIs, mathematical safeguards, parallel merchant processing, and HTML generation, while the analyzed material indicates those behaviors may be absent or mocked. This mismatch is security-relevant because operators and users may grant trust, permissions, or sensitive inputs based on capabilities and safeguards that do not actually exist.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill makes strong claims about real APIs, mathematical safeguards, parallel merchant processing, and HTML generation, while the analyzed material indicates those behaviors may be absent or mocked. This mismatch is security-relevant because operators and users may grant trust, permissions, or sensitive inputs based on capabilities and safeguards that do not actually exist.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The skill makes strong claims about real APIs, mathematical safeguards, parallel merchant processing, and HTML generation, while the analyzed material indicates those behaviors may be absent or mocked. This mismatch is security-relevant because operators and users may grant trust, permissions, or sensitive inputs based on capabilities and safeguards that do not actually exist.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The skill makes strong claims about real APIs, mathematical safeguards, parallel merchant processing, and HTML generation, while the analyzed material indicates those behaviors may be absent or mocked. This mismatch is security-relevant because operators and users may grant trust, permissions, or sensitive inputs based on capabilities and safeguards that do not actually exist.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill makes strong claims about real APIs, mathematical safeguards, parallel merchant processing, and HTML generation, while the analyzed material indicates those behaviors may be absent or mocked. This mismatch is security-relevant because operators and users may grant trust, permissions, or sensitive inputs based on capabilities and safeguards that do not actually exist.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

The skill makes strong claims about real APIs, mathematical safeguards, parallel merchant processing, and HTML generation, while the analyzed material indicates those behaviors may be absent or mocked. This mismatch is security-relevant because operators and users may grant trust, permissions, or sensitive inputs based on capabilities and safeguards that do not actually exist.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The skill makes strong claims about real APIs, mathematical safeguards, parallel merchant processing, and HTML generation, while the analyzed material indicates those behaviors may be absent or mocked. This mismatch is security-relevant because operators and users may grant trust, permissions, or sensitive inputs based on capabilities and safeguards that do not actually exist.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

The skill makes strong claims about real APIs, mathematical safeguards, parallel merchant processing, and HTML generation, while the analyzed material indicates those behaviors may be absent or mocked. This mismatch is security-relevant because operators and users may grant trust, permissions, or sensitive inputs based on capabilities and safeguards that do not actually exist.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The comments state that external API dependencies were removed, yet the engine still creates an external map API client. This mismatch can mislead reviewers and operators into granting trust or deploying the skill under false assumptions, which is dangerous in a security-sensitive agent environment because hidden external integrations can enable unintended data flow or secret use.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The README explicitly advertises use of real third-party travel APIs and describes collecting itinerary details such as destination, budget, party size, and transport preferences, but it provides no user-facing notice, consent flow, or data-sharing disclosure. In a travel-planning skill, those details can reveal sensitive behavioral and personal information, so silent transmission to external services creates a real privacy and transparency risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The README states that the system will 'forcibly send' an HTML guide to the user, but does not describe opt-in, preview, or the outbound behavior's limits. While less severe than credential theft or code execution, undisclosed automatic content delivery can surprise users, create spam-like behavior, and increase the chance of unintended disclosure if the guide contains personalized itinerary data or third-party links.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill declares capabilities that imply shell, environment-variable, and MCP/package execution behavior, but it does not define any explicit tool or permission scope. In a skill ecosystem, that omission can cause overbroad execution authority, make reviewer expectations inaccurate, and increase the chance that the skill accesses sensitive resources or executes commands beyond what users understand.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Broad trigger phrases like generic travel-planning requests can cause the skill to activate in many ordinary conversations without clear user intent. In combination with shell/API/network capabilities and forced output behavior, overbroad activation increases the chance of unintended external calls, data processing, or unsolicited content delivery.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The output metadata states P0-force delivery across multiple channels, but there is no warning that generated HTML may be forcibly sent. Forced multi-channel delivery can create privacy, spam, or content-injection concerns, especially if the HTML includes third-party links or user-derived travel details.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill advertises real external APIs and merchant links but does not present a clear user-facing warning in the relevant section. Users may provide itinerary details, locations, dates, budgets, or other personal travel data without understanding that those details could be sent to third-party services.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

Using npx @fly-ai/flyai-cli without pinning an exact version introduces supply-chain risk because each invocation may resolve to a different package version. If the upstream package or dependency chain is compromised, the skill could execute attacker-controlled code at runtime.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The usage section repeats vague activation wording and still does not define boundaries for what the skill will do when triggered. This can mislead users into invoking a capability set that includes network access and commercial link generation without sufficiently informed consent.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
72% confidence
Finding

The documented nohup ... & watchdog launch establishes a persistent background process. Persistence is not always malicious, but in a skill context it increases operational risk by enabling long-lived unattended execution, which can continue making external calls, handling data, or evading normal session boundaries.

Content

Scanner excerpt · SKILL.md (reported line 176)May include surrounding context.

7.2 守护启动

bash
nohup bash watchdog.sh > /tmp/watchdog.log 2>&1 &

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The description is presented entirely in Chinese and the manifest provides no indication that users can choose another language or that the skill is intentionally limited to a Chinese-speaking or region-specific context. This creates a natural-language policy concern because it appears to impose a locale without opt-in or documented justification.

Content

No source excerpt is available for this finding.