T08 · Insecure Dependencies
- Location
reference/core-workflow.md:5- Finding
Unpinned Global Dependency Installation with Optional Root Privileges
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill is a coherent travel assistant, but it tells agents to use unsafe network and installation practices and to persist detailed travel profiles with inconsistent consent controls.
Review carefully before installing. Do not use the documented TLS-bypass commands, avoid sudo or global @latest installs, and only allow profile saving after explicit confirmation. If used, prefer a pinned, trusted CLI installed without elevated privileges and keep or delete the local travel profile deliberately.
reference/core-workflow.md:5Unpinned Global Dependency Installation with Optional Root Privileges
reference/core-workflow.md:57TLS Certificate Verification Disabled for FlyAI Network Requests
reference/core-workflow.md:57Shell Command Injection Through User-Controlled Travel Parameters
reference/memory-system.md:50Persistent User Profiling Without Consistent Explicit Consent
The trigger phrases are extremely broad (旅行, 度假, 去哪玩, etc.), making accidental invocation likely during ordinary conversation. Overbroad activation increases the chance that memory reads, profile access, or persistent preference storage occur without the user clearly intending to invoke this skill.
The skill explicitly instructs disabling TLS certificate verification with NODE_TLS_REJECT_UNAUTHORIZED=0 when SSL validation fails. This removes transport security checks and enables man-in-the-middle interception or tampering of fetched travel data, which is especially risky because the skill relies on external content and search results for planning decisions.
The workflow disables TLS certificate verification for FlyAI searches without any warning, normalizing an unsafe network posture. This enables traffic interception and response tampering, which is especially risky because returned data includes travel options and jumpUrl booking links that users may trust and follow.
The workflow explicitly sets NODE_TLS_REJECT_UNAUTHORIZED=0 before networked FlyAI searches, which disables TLS certificate validation and makes HTTPS connections vulnerable to man-in-the-middle interception or tampering. In a travel assistant context, these commands may handle itinerary, booking, and preference data, so the unsafe setting can expose sensitive user information and allow manipulated search results or malicious links.
The documentation explicitly instructs users to set NODE_TLS_REJECT_UNAUTHORIZED=0, which disables TLS certificate validation for all HTTPS connections made by the Node.js process. This makes man-in-the-middle interception possible and can expose travel queries, account/session data, or backend responses to tampering if the command runs on an untrusted network or against a compromised proxy.
The file normalizes bypassing TLS checks as required usage without clearly warning that it disables server identity verification and transport integrity. Because this is command reference material, users are likely to copy-paste it broadly, turning an insecure troubleshooting step into a standing operating procedure.
The documentation explicitly instructs operators to set NODE_TLS_REJECT_UNAUTHORIZED=0 for all FlyAI commands, which disables TLS certificate verification globally for those invocations. This allows man-in-the-middle interception or spoofing of travel search/API responses, exposing user queries and enabling tampered pricing, booking, or itinerary data to be returned without detection.
The skill tells operators to disable TLS certificate validation without any adequate safety boundary, making insecure transport the expected mode of operation. In a travel-planning context that handles destinations, dates, budgets, and possibly booking-related links, this creates a realistic avenue for traffic interception and response tampering.
The fallback mode writes detailed travel-profile data to a predictable file in the user's home directory (~/.flyai/user-profile.md), which can expose sensitive personal information to other local processes, users, backups, or synchronization tools. Storing family details, budget, travel history, and special needs in cleartext on disk materially increases privacy and confidentiality risk beyond the stated travel-assistant purpose.
The skill states that it will read, remember, and update user preferences persistently at startup and during conversation, but it does not require a clear upfront privacy notice or affirmative consent before doing so. This creates a privacy risk because sensitive travel habits, home city, companions, and preferences may be collected and reused unexpectedly.
The fallback behavior reads a local file at ~/.flyai/user-profile.md without a clear warning or explicit consent at the point of access. Accessing local profile data silently is risky because it expands the skill's reach into user-resident data and may expose personal information unrelated to the immediate request.
Setting NODE_TLS_REJECT_UNAUTHORIZED=0 establishes an unsafe default by globally disabling certificate validation for Node.js HTTPS connections. In the context of a travel assistant that may fetch live pricing, booking-related information, or external content, this materially increases the risk of spoofed responses, data interception, and downstream misinformation.
| 用户多轮后改主意 | 支持随时回退:"行,换个方向重新搜" |
| 搜索返回空结果 | 降级处理,给出文字建议 + 手动搜索入口 |
| 用户停留灵感阶段 | 不强推预订,保存方案供后续查看 |
| SSL 证书验证失败 | 确保命令前加 `NODE_TLS_REJECT_UNAUTHORIZED=0` |
| 用户提及专项能力关键词 | 自动路由到对应专项能力执行 |
| **需求超出预设场景** | **启动自主探索模式,组合能力解决** |
The skill instructs users to install or upgrade a CLI globally and suggests elevated privileges, but provides no warning about the security implications of system-wide package installation or execution of package scripts. This creates unnecessary supply-chain and host-integrity risk, especially for an assistant whose primary role is travel planning rather than local software administration.
The presence of sudo in installation guidance instructs users to run a network-fetched package installation with root privileges. If the package, registry path, or installation scripts are compromised, the attacker gains elevated control over the host; this is disproportionate to the needs of a travel assistant skill.
| 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,确保可以访问 registry.npmjs.org |
### 阶段1: 需求发散 (Discovery)
The workflow recommends using sudo for a global npm install of the FlyAI CLI, which elevates privileges for downloading and executing package install scripts from the network. This increases the blast radius of any package compromise, typo-squatting, or supply-chain issue and is unjustified for a normal travel-planning workflow.
The workflow explicitly tells the agent to extract multiple categories of user-provided information, including travel plans, companions, budget, and special needs, and later reuse them in searches and generated outputs. Although framed as normal product behavior, this is a natural-language instruction to persist and propagate potentially sensitive user data across steps, which can create data-leak risk if the skill is reused in broader contexts.
Using NODE_TLS_REJECT_UNAUTHORIZED=0 as a default for keyword search establishes an insecure baseline by bypassing certificate checks. This makes the command susceptible to MITM attacks and malicious result injection, undermining the integrity and confidentiality of user queries and returned recommendations.
基于收集到的信息,调用 FlyAI 搜索:
NODE_TLS_REJECT_UNAUTHORIZED=0 flyai keyword-search --query "[用户描述关键词]"
输出候选目的地,格式:
Disabling TLS validation for flight search exposes origin, destination, and date data in a channel where responses can be intercepted or modified. Because these results may influence bookings and travel decisions, manipulated responses could redirect users to fraudulent or misleading options.
# 搜索机票
NODE_TLS_REJECT_UNAUTHORIZED=0 flyai search-flight \
--origin "[出发城市]" --destination "[目的地]" \
--dep-date [出发日期] --back-date [返回日期] --sort-type 3
Disabling TLS verification for hotel search creates the same network-layer weakness for lodging queries and returned booking data. In this context, tampered responses could alter pricing, destinations, or booking links, increasing fraud and privacy risks.
--dep-date [出发日期] --back-date [返回日期] --sort-type 3
# 搜索酒店
NODE_TLS_REJECT_UNAUTHORIZED=0 flyai search-hotel \
--dest-name "[目的地]" --check-in-date [入住日期] \
--check-out-date [退房日期] --sort rate_desc
Using an unsafe TLS default for POI search permits interception and tampering with search responses, which can mislead itinerary generation and expose user travel interests. Repeating this pattern across all network commands shows a systemic insecure default rather than an isolated mistake.
--check-out-date [退房日期] --sort rate_desc
NODE_TLS_REJECT_UNAUTHORIZED=0 flyai search-poi --city-name "[目的地]"
**2.3 输出初步方案**
This markdown file contains natural-language instructions only in Chinese, which can amount to forcing a specific language on users without opt-in. The policy requires flagging language or locale constraints unless the skill offers a choice or clearly documents and justifies the restriction.
Using NODE_TLS_REJECT_UNAUTHORIZED=0 is an unsafe default because it disables certificate validation process-wide for the invoked command. In a travel assistant context, network responses may include search results, booking-related data, or authenticated API traffic that an attacker could intercept or modify.
> ⚠️ SSL 证书问题:执行 flyai 命令时,需在命令前添加环境变量绕过 SSL 验证:
> ```bash
> NODE_TLS_REJECT_UNAUTHORIZED=0 flyai <command>
> ```
## fliggy-fast-search
This example embeds TLS verification bypass directly into the fliggy-fast-search command, encouraging insecure copy-paste use for routine searches. Even seemingly low-sensitivity search traffic can be leveraged for response tampering, phishing links, tracking, or theft of tokens/cookies if present in the CLI environment.
自然语言全品类搜索,一次查询覆盖机票、酒店、门票、旅游等所有品类。
NODE_TLS_REJECT_UNAUTHORIZED=0 flyai fliggy-fast-search --query "[自然语言查询]"
示例:
The search-flight example hardcodes a process-wide TLS bypass for flight queries, making integrity of pricing, routing, and availability data untrustworthy. An attacker positioned on the network could alter results or inject deceptive booking information, which is especially risky in a travel-planning workflow where users may act on returned options.
结构化机票搜索,支持深度筛选。
NODE_TLS_REJECT_UNAUTHORIZED=0 flyai search-flight \
--origin "[出发城市]" \
--destination "[目的地]" \
--dep-date [出发日期] \
The hotel search example repeats the insecure default, so users performing accommodation searches may unknowingly accept unauthenticated HTTPS connections. This enables interception or manipulation of hotel listings, pricing, and related metadata, undermining trust and potentially facilitating fraud.
结构化酒店搜索,支持星级、价格、位置筛选。
NODE_TLS_REJECT_UNAUTHORIZED=0 flyai search-hotels \
--dest-name "[目的地]" \
--check-in-date [入住日期] \
--check-out-date [退房日期]