Back to skill

Security audit

中转不浪费

Security checks for vulnerabilities and agentic risk

Overview

The skill is a real travel-planning assistant, but it asks the agent to install mutable global software, bypass TLS checks, add unrequested booking links, and persist personal travel profile data.

Review before installing. This skill may be useful for Chinese-language layover travel planning, but users should not allow sudo/global runtime installs, should not bypass TLS certificate validation, and should verify any booking links before clicking. Treat profile storage as opt-in only and avoid saving sensitive travel or family details unless you understand where they will persist.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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 (3)

T01 · Skill Instruction Hijacking

Warning
Location
reference/workflow.md:146
Finding

Mandatory Injection of Unrequested Commercial Searches and Booking Links

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Error
Location
reference/workflow.md:3
Finding

Mutable Global Dependency Installation with Optional Root Privileges

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
reference/workflow.md:69
Finding

TLS Certificate Verification Disabled for FlyAI Commands

Content
View full analysis
``` ``` ### Technical Analysis Setting `NODE_TLS_REJECT_UNAUTHORIZED=0` disables TLS certificate validation for the Node.js process. Encryption may still be used, but the client no longer verifies that it is communicating with the intended server. This converts a certificate error—which should cause a fail-closed response—into an unauthenticated network connection. Any attacker capable of intercepting traffic can impersonate the remote service and modify responses. The risk is amplified because the workflow consumes travel prices, schedules, and booking URLs from the service and later displays returned links as trusted booking actions. A man-in-the-middle attacker could therefore inject false travel data or malicious links into the final answer. ### Attack Path 1. The FlyAI command encounters a certificate-chain validation error. 2. The workflow instructs the agent or user to rerun it with `NODE_TLS_REJECT_UNAUTHORIZED=0`. 3. An attacker on the local network, compromised proxy, hostile Wi-Fi network, or upstream routing path intercepts the TLS connection. 4. Because certificate validation is disabled, the attacker presents an arbitrary certificate without rejection. 5. The attacker returns modified flight, hotel, attraction, price, or booking-link data. 6. The Skill incorporates the manipulated data into its recommendations. 7. The user may follow an attacker-controlled booking link, disclose information, or make decisio ...[truncated 645 chars]
Remediation
View remediation
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
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Rogue AgentSelf-Modification, Session Persistence
Findings (24)

Natural-Language Policy Violations

High
Category
Not specified by scanner
Confidence
87% confidence
Finding

L03 的整体描述明确面向中文用户并特别提到“对华签证政策”,但未说明这是区域专用技能,也未提供其他语言或用户选择机制。按照语言/locale 政策要求,若技能限制特定语言或地区,应有明确的选择或合理、文档化的约束说明。

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The referenced content documents a train-search capability, but the skill is described as a flight layover-tour assistant. This kind of capability/reference mismatch can cause the agent to invoke the wrong tool, mis-handle user travel requests, or expose unintended functionality outside the stated scope. In a travel-planning skill, incorrect modality assumptions can lead to unsafe or materially wrong itinerary advice, especially when the skill is supposed to calculate tight layover feasibility and time buffers.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The workflow instructs operators to disable TLS certificate verification with NODE_TLS_REJECT_UNAUTHORIZED=0 when SSL errors occur. This removes server identity validation and makes all subsequent CLI traffic vulnerable to man-in-the-middle interception or tampering, which is especially dangerous in a workflow that retrieves booking links and travel data.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

Telling users to bypass certificate validation without an explicit warning normalizes an unsafe troubleshooting pattern. In practice, this can expose flight search queries, returned URLs, and any future authenticated API traffic to spoofed endpoints and modified responses.

Content

No source excerpt is available for this finding.

Ssd 2

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The instruction explicitly suggests bypassing SSL validation failures rather than resolving their root cause. That creates an insecure default operating mode where tampered or spoofed FlyAI responses could be trusted and surfaced to users as legitimate booking or travel guidance.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

L03 将触发条件写成“当用户提到……”并列出多个短语,其中“转机时间长”“便宜的中转”等表述较宽泛,可能出现在一般旅行讨论中,而不仅限于需要本技能的场景。描述中也未提供排除条件或负例,难以判断何时应激活、何时不应激活。

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill instructs itself to read user history and, on fallback, a local profile file without an explicit user-facing consent or notice at the point of access. That creates a privacy risk because the skill may access personal preference data unexpectedly, increasing the chance of over-collection or disclosure of sensitive travel-related habits.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

SQP-3 applies to all file types and includes language or locale policy violations. This file presents the interaction exclusively in Chinese and does not indicate that language selection is user-configurable or that the skill is intentionally limited to a Chinese-language or region-specific context.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This markdown file uses a fixed Chinese currency label at L21 and all examples/output are presented in Chinese place names and localized fields, but the document does not state that the skill is China-specific or provide any language/locale opt-in. That creates a natural-language locale policy concern because users are implicitly constrained to a specific language and regional format.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The document expands a layover-planning skill into persistent cross-session profile storage, including personal travel preferences and household details, without showing that such retention is necessary for the immediate task. This creates unnecessary data accumulation and increases privacy risk if the stored profile is later accessed, reused, or exposed beyond the user's expectation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The document instructs the skill to store user profile data in Qoder Memory or a local file but provides no warning about persistence, local exposure, consent, or handling risks. Omitting these safeguards makes accidental overcollection and silent retention more likely, especially in environments where users may not expect local filesystem storage.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
89% confidence
Finding

Creating and using ~/.flyai/user-profile.md establishes session persistence on the local filesystem for personal profile data, which can remain accessible to other local processes, users, backups, or later sessions. In the context of a transit-planning skill, this persistence is more dangerous because it is not essential to core functionality and may occur in environments without strong access controls or clear user awareness.

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
96% confidence
Finding

The stored schema includes detailed personal attributes such as home city, airport, budget, party size, presence of a child, past destinations, and special needs, which exceed what is clearly required for generating a one-off transit-tour suggestion. Persisting this level of profiling can enable sensitive inference about the user and magnifies harm if the data is mishandled or retained indefinitely.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

The file consistently mandates workflow and user-facing prompts in Chinese, including required question examples, but does not indicate that the user may choose another language. A skill that effectively forces a specific language without opt-in can violate language/locale policy requirements.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
85% confidence
Finding

Recommending sudo npm install -g runs a network-fetched package installation with elevated privileges. If the package, dependency chain, registry path, or local npm execution flow is compromised, this can lead to full system-level code execution and persistence.

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

NODE_TLS_REJECT_UNAUTHORIZED=0 is an unsafe default because it globally disables certificate verification for the invoked Node.js process. In this skill context, that means a transient certificate issue can silently downgrade transport security for flight-search and booking-link retrieval, enabling interception and response manipulation.

Content

Scanner excerpt · reference/workflow.md (reported line 74)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" 或类似 UNABLE_TO_VERIFY_LEAF_SIGNATURE 错误,需要在命令前加上环境变量:

bash
NODE_TLS_REJECT_UNAUTHORIZED=0 flyai <command>

2.1 搜索直飞航班(作为对比基准)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The manifest describes a helper focused on finding long-layover flight options, evaluating whether a transit city can be visited, and generating a half-day transit itinerary with time-safety calculations. These lines add destination hotel search and destination attraction ticket search, which are separate travel-booking capabilities not necessary to evaluate or plan a layover stopover.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest presents the skill as an assistant for identifying suitable 6-15 hour layovers and generating a safe transit-city half-day plan. This output template requires surfacing flight, hotel, and attraction booking links plus bundled purchase suggestions, shifting the skill from analysis/recommendation into a broader booking funnel not described in the manifest.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The only natural-language examples shown are Chinese queries, which can imply the skill expects or prefers a specific language. There is no accompanying note that other languages are supported or that Chinese is required for a justified regional use case.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

This markdown file presents all instructions and recommendations exclusively in Chinese. Under the policy for natural-language violations, forcing a specific language without offering user choice or documenting a justified locale constraint can be a policy issue.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The example explicitly shows external flight and POI searches, which implies user travel details may be sent to third-party services, but it provides no notice, consent step, or minimization guidance. While the data is not highly sensitive in most cases, travel plans can still reveal location, timing, and behavioral information, so the omission creates a real privacy/transparency issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The parameter documentation requires specific Chinese values such as 酒店, 民宿, 客栈, 大床房, and 双床房, which imposes a language-specific interface. The file does not explain that this skill is intended only for a Chinese travel service or otherwise justify the locale constraint.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The file presents accepted values and usage examples in Chinese only, such as bed-type values and destination/hotel examples, with no indication that other languages or locales are supported. This can be a natural-language policy concern if skills are expected not to force a specific language without user opt-in.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

This markdown file is in scope for natural-language policy checks. The examples and enumerated values are effectively Chinese-only, which can amount to forcing a specific language or locale unless the skill explicitly offers user opt-in or explains that it is region-specific.

Content

No source excerpt is available for this finding.