Back to skill

Security audit

周末去哪儿

Security checks for vulnerabilities and agentic risk

Overview

This travel planner is not clearly malicious, but it needs review because it directs risky system setup, unsafe network troubleshooting, and persistent handling of personal travel profile data.

Review before installing. Use this only if you are comfortable with a skill that may install a global FlyAI CLI, contact external travel services, and use stored travel preferences. Do not run the documented `sudo npm install` fallback or disable TLS verification; prefer a pinned, user-scoped install and require explicit consent before reading or saving any profile data.

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 (5)

T08 · Insecure Dependencies

Error
Location
reference/workflow.md:3
Finding

Unpinned Global Dependency Installation with Privileged Fallback

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
reference/workflow.md:70
Finding

TLS Certificate Verification Can Be Disabled for FlyAI Requests

Content
View full analysis
``` ### Technical Analysis Setting `NODE_TLS_REJECT_UNAUTHORIZED=0` disables TLS certificate validation for the complete Node.js process running the FlyAI CLI. The client will no longer reliably verify that it is communicating with the intended server. This is a fail-open response to certificate errors. It permits connections using expired, self-signed, untrusted, or attacker-supplied certificates instead of correcting the certificate or trust-store problem. Because the CLI retrieves travel results and booking URLs, a man-in-the-middle attacker could alter prices, availability, destination information, or links returned to the user. ### Attack Path 1. A user invokes a FlyAI search over an untrusted or attacker-controlled network. 2. A certificate verification error occurs naturally or is deliberately induced. 3. The workflow retries the command with `NODE_TLS_REJECT_UNAUTHORIZED=0`. 4. The attacker intercepts the TLS connection and presents an untrusted certificate. 5. The CLI accepts the certificate because validation has been disabled. 6. The attacker observes or modifies requests and responses. 7. Manipulated booking links or travel results are presented as legitimate Skill output. ### Impact Assessment An attacker in a suitable network position can compromise the confidentiality and integrity of data transmitted by the CLI. This may expose travel dates, departure locations, destinations, budgets, and preferences. Response manipulation could redirect users to phishing or fraudulent booking pages, present falsified prices, or influence purchases. This finding does not independently grant local system privileges, but it compromises all data exchanged by the affected process. ]]>
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
reference/workflow.md:80
Finding

User-Controlled Values Are Interpolated into Shell Command Templates

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
reference/user-profile-storage.md:63
Finding

Sensitive Travel and Family Profile Data Is Persisted in Plaintext

Content
View full analysis
Last updated: 2026-04-03 15:30 ## Basic Information - Residence city: Hangzhou - Departure airport: Xiaoshan Airport ## Travel Preferences - Budget preference: Medium - Typical travelers: 2 - Family members: Child, age 3 - Preferred travel types: Islands, family travel, natural scenery - Accommodation preference: Four stars or above ## Travel History - Previously visited cities: Sanya, Xiamen, Dali ## Special Requirements - Pet friendly: No - Accessibility requirement: No ``` The profile is stored at: ```text ~/.flyai/user-profile.md ``` ### Technical Analysis The documented local-storage mode persists profile data as a Markdown file in the user's home directory. The data model includes residence, preferred airport, budget, family composition, a child's age, travel history, and accessibility requirements. The workflow appropriately requests confirmation before writing newly discovered preferences. However, it does not specify restrictive directory or file permissions, encryption, data minimization, a retention period, deletion controls, or secure update behavior. Default permissions depend on the host's umask and environment. Backups, synchronization tools, malware, other local processes, or other users on a poorly configured multi-user system may therefore access the profile. ### Attack Path 1. The Skill identifies a new travel preference or personal attribute. 2. The user approves persistent storage. 3. The Skill creates `~/.flyai` and writes `user-profile.md` using ambient default permissions. 4. Sensitive profile fields remain stored indefinitely in plaintext. 5. Another local process, user, backup system, or compromised development tool reads the file. 6. The expos ...[truncated 552 chars]
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
SKILL.md:22
Finding

Persistent User Profile Is Read Automatically Without Session-Specific Consent

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • 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
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
Findings (20)

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md
---
name: flyai-weekend-trip
description: 周末去哪儿2天1夜说走就走方案助手。输入出发城市和本周末日期,一次性生成3个"说走就走"的周末方案,每个包含目的地+航班/高铁+酒店+精选景点+总价估算。当用户提到"周末去哪"、"周末出游"、"2天1夜"、"说走就走"、"周末旅行"、"这周末出去玩"、"短途游"时使用。
---

# 周末去哪儿 - 2天1夜说走就走方案

周末快速出游方案规划师,帮助用户从"想出去"到"订好出发"快速决策周末短途旅行。

## FlyAI 能力

> 完整命令参考见 reference 目录

**主�

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The workflow recommends sudo npm install -g as a troubleshooting step, encouraging elevated execution of code fetched from a package registry. Running package installation with root privileges magnifies the impact of any malicious package, compromised dependency, install script, or typo-squatting incident into full system compromise.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The workflow explicitly instructs users to disable TLS certificate verification with NODE_TLS_REJECT_UNAUTHORIZED=0 to work around SSL failures. This removes server identity validation and exposes all subsequent FlyAI CLI network traffic to man-in-the-middle interception or tampering, which is especially dangerous because the skill also relies on package installation and travel search results from network services.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

This instruction tells users to disable TLS certificate verification for CLI commands and does so without any warning about the consequences. In practice it normalizes insecure transport behavior and can let an attacker intercept, alter, or spoof responses from remote services that the skill depends on.

Content

No source excerpt is available for this finding.

Ssd 1

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill recommends bypassing SSL errors by disabling certificate verification instead of correcting the root cause. This undermines transport security and makes the user vulnerable to malicious network intermediaries serving manipulated travel data or potentially weaponized package/content responses.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases are broad, generic travel intents such as '周末去哪' and '短途游', which can cause the skill to activate in many ordinary travel conversations without clear user intent to use this specific capability. Overbroad activation increases the chance the skill runs unexpectedly and accesses profile-related data or pushes users into booking-oriented flows when a simpler answer was expected.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill is designed to read stored user profile data at startup, but the user-facing description and activation behavior do not clearly warn users that profile or local file data may be accessed automatically. This creates a transparency and privacy risk because the skill may process personal preference data without informed user awareness at the moment of invocation.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest describes a narrowly scoped assistant for generating 3 weekend 2-day/1-night getaway plans based on departure city and this weekend’s date. This reference file documents a generic 'ai-search' capability supporting hotels, attractions, flights, trains, mixed travel intent, complex constraints, and even non-weekend multi-day international trips, which is materially broader than the declared skill purpose.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The file requires specific Chinese bed-type values and uses only Chinese destination, brand, and hotel-name examples, which indicates a fixed language/locale expectation. There is no note that the skill is China-specific or that users may provide other locales/languages, creating a natural-language locale policy concern.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The document defines persistent cross-session storage of a detailed user travel profile even though the skill is described as a weekend trip planner, not a profile management system. This creates unnecessary data collection and retention, expanding the skill's access to personal information beyond what is required for the immediate task and increasing privacy risk if the data is later exposed or reused unexpectedly.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The guidance instructs reading and writing user profile data to Qoder Memory and a local file but does not include privacy notices, sensitivity classification, consent requirements, or security expectations for stored data. This omission can lead to silent persistence of personal data and inconsistent protection across platforms, especially when local files are used.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
84% confidence
Finding

Creating and using ~/.flyai/user-profile.md establishes session persistence on the local machine, allowing personal preference data to survive beyond the current interaction. In this skill context, persistence is more dangerous because it stores identifiable travel and household preference information without any stated access controls, retention policy, or cleanup mechanism.

Content

Scanner excerpt · reference/user-profile-storage.md (reported line 73)May include surrounding context.

如果 ~/.flyai/ 目录不存在,需要先创建:

bash
mkdir -p ~/.flyai

文件格式

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The specified schema stores long-term attributes such as home city, airport, budget, family composition, accommodation preferences, trip history, and special needs. For a lightweight '2-day/1-night weekend trip' assistant, this is excessive profiling and can reveal sensitive lifestyle and household information that is not strictly necessary for one-off itinerary generation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The workflow directs a global npm install/upgrade of a CLI as a mandatory first step without warning that it changes the local system environment. This can unexpectedly modify PATH-available tooling, introduce supply-chain risk, and violate the principle of least surprise for a skill that appears to be a travel-planning assistant rather than a system-management tool.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
96% confidence
Finding

The presence of sudo in an installation command is a real privilege-escalation risk because it invites root execution of externally sourced package installation logic. In the context of a consumer travel skill, this is unnecessary and disproportionately dangerous relative to the feature being provided.

Content

Scanner excerpt · reference/workflow.md (reported line 20)May include surrounding context.

md
| 情况 | 处理方式 |
|-----|---------|
| npm 未安装 | 提示用户先安装 Node.js (https://nodejs.org/) |
| 权限不足 | 建议使用 `sudo npm install -g @fly-ai/flyai-cli@latest --registry=https://registry.npmjs.org` 或使用 nvm 管理 Node |
| 网络问题 | 建议用户检查网络或使用国内镜像 `npm config set registry https://registry.npmmirror.com` |

**注意:** 此步骤只在首次使用时执行,后续调用会直接跳过已安装的情况。

Unsafe Defaults

Medium
Category
Tool Misuse
Confidence
99% confidence
Finding

Setting NODE_TLS_REJECT_UNAUTHORIZED=0 creates an unsafe default for the invoked command by globally disabling Node's TLS certificate validation. This materially lowers the security baseline and can turn ordinary network errors into silent insecure connections that users may trust.

Content

Scanner excerpt · reference/workflow.md (reported line 77)May include surrounding context.

重要:SSL 证书验证问题处理 如果遇到 "SSL 证书验证失败" 错误,需要在命令前加上环境变量:

bash
NODE_TLS_REJECT_UNAUTHORIZED=0 flyai <command>

2.1 自然语言广泛搜索确定候选目的地

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

The only natural-language query examples are in Chinese, which can imply the skill is intended to be used in a specific language without any stated user choice or justification. Because this is a reference markdown file and no alternative language or opt-in is mentioned, it may conflict with language/locale neutrality expectations.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

This markdown file presents currency as CNY and all example inputs/outputs in Chinese, including city names, weekday abbreviations, and cabin/journey labels, without stating that the skill is China-specific or that users may choose another language/locale. That creates a natural-language locale policy concern because the documentation appears to assume a fixed language/region by default.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The markdown specifies accepted values such as 酒店, 民宿, 客栈, 大床房, and 双床房, and the examples use Chinese destinations, but it does not state that the skill is intentionally China-specific or offer an alternative language/locale. This can amount to a natural-language locale constraint that is not documented as an opt-in or justified regional limitation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

The skill documentation presents keyword and category inputs only in Chinese, and the allowed category values are entirely Chinese-language terms. Without any note that the skill is China-specific or that other languages are unsupported, this creates an implicit locale constraint that may violate language/locale policy expectations.

Content

No source excerpt is available for this finding.