Back to skill

Security audit

签证时间线

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a visa-planning assistant, but it asks users to install mutable global software, sometimes with sudo, disables TLS checks, and uses broad persistent travel-profile storage.

Install only if you are comfortable with this skill modifying your Node environment and using persistent travel-profile data. Do not run the sudo install or TLS-bypass command; prefer a pinned, reviewed, user-local CLI setup and delete or narrow the unrelated travel-search references and stored-profile behavior before routine use.

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 (4)

T08 · Insecure Dependencies

Error
Location
reference/core-workflow.md:3
Finding

Mandatory Installation of an Unpinned Global Dependency

Content
View full analysis

Vulnerability Details

File Location: reference/core-workflow.md:3-8
Vulnerability Type: Unpinned global package installation and supply-chain exposure
Risk Level: High

bash
npm install -g @fly-ai/flyai-cli@latest --registry=https://registry.npmjs.org

Technical Analysis

The workflow requires installing or upgrading @fly-ai/flyai-cli before performing a search. The mutable latest tag is used instead of an audited exact version, and no package integrity digest or lockfile is provided.

Consequently, the code installed when the Skill runs may differ from the code that existed when the Skill was audited. npm packages can contain installation lifecycle scripts, such as preinstall, install, and postinstall, which execute during installation. A compromised publisher account, malicious package release, or upstream package compromise could therefore introduce arbitrary executable behavior.

Global installation also modifies the user's shared executable environment rather than isolating the dependency to this Skill.

Attack Path

  1. An attacker compromises the npm publisher, package, or release pipeline for @fly-ai/flyai-cli.
  2. The attacker publishes a malicious release and assigns it the latest tag.
  3. A user invokes a workflow that requires a FlyAI search.
  4. The Skill directs the environment to run the global installation command.
  5. npm downloads the attacker-controlled release.
  6. Malicious lifecycle scripts execute during installation, or the installed flyai executable runs malicious code when subsequently invoked.

Impact Assessment

Successful exploitation permits arbitrary code execution with the privileges of the account running npm. The malicious package could read or modify user-accessible files, access environment variables and credentials available to that account, tamper with globally installed commands, or make unauthorized network requests.

The impact becomes s ...[truncated 90 chars]

Remediation
View remediation

Remediation Suggestions

  • Replace @latest with one audited, exact package version.
  • Record and verify the package integrity digest.
  • Use a lockfile and a local project dependency rather than a global installation.
  • Do not automatically upgrade the dependency whenever the Skill is invoked.
  • Require explicit user approval before installing any external software.
  • Disable npm lifecycle scripts where compatible with the package.
  • Run the dependency in a restricted sandbox or container with minimal filesystem and network access.
  • Establish a controlled dependency-update process that performs security review before changing the pinned version.

T05 · Unauthorized Access and Privilege Escalation

Error
Location
reference/core-workflow.md:16
Finding

Privilege Escalation Through Recommended Root-Level Package Installation

Content
View full analysis

Vulnerability Details

File Location: reference/core-workflow.md:16-20
Vulnerability Type: Unnecessary privileged execution of an external package installer
Risk Level: Critical

bash
sudo npm install -g @fly-ai/flyai-cli@latest --registry=https://registry.npmjs.org

Technical Analysis

When the ordinary global installation encounters a permission error, the workflow recommends repeating it through sudo. Visa planning and travel searches do not legitimately require administrative access.

This command combines root privileges with an unpinned third-party dependency. npm installation lifecycle scripts may execute during package installation. If the selected package release or one of its dependencies is compromised, attacker-controlled code may consequently execute as root.

The recommendation violates least privilege and converts a dependency compromise that would otherwise be limited to a user account into a potential full-system compromise.

Attack Path

  1. An attacker introduces malicious code into the package selected by the mutable latest tag or one of its transitive dependencies.
  2. A user attempts the mandatory global installation.
  3. The installation fails because the user cannot modify the global npm directory.
  4. Following the documented recovery procedure, the user runs the command with sudo.
  5. npm downloads the malicious package and executes its installation logic with root privileges.
  6. The attacker modifies protected system files, installs privileged executables, accesses root-readable data, or creates additional persistence mechanisms.

Impact Assessment

Successful exploitation can provide root-level arbitrary code execution and complete host compromise. Potential effects include reading or modifying any local file, stealing credentials, changing system configuration, replacing trusted executables, disabling security controls, creating privileged accounts, and install ...[truncated 281 chars]

Remediation
View remediation

Remediation Suggestions

  • Remove the sudo npm install recommendation entirely.
  • Install dependencies within a user-owned project directory.
  • Use a user-scoped Node.js installation managed through tools such as nvm.
  • Pin and review the exact dependency version before installation.
  • Run third-party tooling under a dedicated unprivileged account or inside a sandbox.
  • Restrict filesystem and network access to the minimum required for travel searches.
  • Treat permission failures as a configuration issue rather than a reason to elevate privileges.
  • If administrative installation is unavoidable, require a separate, explicit security review and prohibit package lifecycle scripts.

T09 · Insecure Skill Coding Practices

Error
Location
reference/core-workflow.md:110
Finding

TLS Certificate Verification Disabled for FlyAI Requests

Content
View full analysis

Vulnerability Details

File Location: reference/core-workflow.md:110-112
Vulnerability Type: Disabled TLS peer authentication
Risk Level: High

bash
NODE_TLS_REJECT_UNAUTHORIZED=0 flyai keyword-search --query "{destination country} visa"

Technical Analysis

NODE_TLS_REJECT_UNAUTHORIZED=0 disables TLS certificate verification for HTTPS requests made by the Node.js process. The client can no longer reliably authenticate the remote endpoint, allowing invalid, self-signed, expired, or attacker-provided certificates to be accepted.

This configuration is supplied specifically as a workaround for certificate-chain failures. Certificate validation errors should instead be resolved by repairing the local trust store or configuring an authorized certificate authority. Disabling validation transforms an observable connection failure into a silent man-in-the-middle exposure.

Attack Path

  1. The user encounters an UNABLE_TO_GET_ISSUER_CERT_LOCALLY error.
  2. The user or agent follows the documented fallback and sets NODE_TLS_REJECT_UNAUTHORIZED=0.
  3. An attacker with a network interception position presents an untrusted TLS certificate.
  4. The Node.js client accepts the certificate because verification is disabled.
  5. The attacker observes or modifies requests and responses.
  6. Altered visa information, prices, or service links are returned to the agent and may be presented to the user as legitimate results.

Impact Assessment

A successful man-in-the-middle attack may expose travel-search parameters and other data transmitted by the CLI. It can also permit response manipulation, including falsified visa guidance, malicious product links, altered pricing, or misleading travel results.

The scope includes all TLS traffic made by the affected Node.js process while the environment variable is active. The setting does not itself grant local administrative privileges, but manipulated responses may su ...[truncated 48 chars]

Remediation
View remediation

Remediation Suggestions

  • Remove the NODE_TLS_REJECT_UNAUTHORIZED=0 workaround.
  • Fail closed when the server certificate cannot be validated.
  • Repair the host operating system or Node.js CA trust store.
  • Where an authorized enterprise proxy is used, install only its approved CA certificate and scope that trust appropriately.
  • Keep certificate and hostname validation enabled.
  • Add clear diagnostics for expired certificates, missing intermediates, and proxy-related trust failures.
  • Avoid returning product links or visa information from a connection whose endpoint identity cannot be authenticated.

T08 · Insecure Dependencies

Error
Location
reference/tools.md:26
Finding

Conflicting and Unpinned Package Identity Executed Through npx

Content
View full analysis

Vulnerability Details

File Location: reference/tools.md:26-28
Vulnerability Type: Dependency identity inconsistency and mutable runtime package execution
Risk Level: High

bash
npx @anthropic-ai/flyai-cli@latest

Technical Analysis

This documentation identifies @anthropic-ai/flyai-cli as the package used to access travel data, while reference/core-workflow.md directs installation of @fly-ai/flyai-cli. The project provides no explanation or provenance establishing why two different npm namespaces are used.

In addition, npx with the mutable latest tag may download and execute package code at runtime. The effective executable can therefore change after the Skill is audited. The inconsistent identity increases the risk that an agent follows the wrong instruction, invokes an unintended package, or is exposed to namespace confusion or package spoofing.

The audit evidence does not establish that either named package is currently malicious. The vulnerability is the unsafe, inconsistent, and unpinned dependency execution mechanism.

Attack Path

  1. An agent follows reference/tools.md rather than the conflicting package instruction in reference/core-workflow.md.
  2. npx resolves @anthropic-ai/flyai-cli@latest from the configured npm registry.
  3. An attacker compromises that package, its publisher, or its release pipeline, or exploits confusion caused by the inconsistent package identity.
  4. A malicious release becomes the version selected by latest.
  5. npx downloads and executes the attacker-controlled package.
  6. The package performs unauthorized actions with the invoking user's permissions.

Impact Assessment

Successful exploitation can result in arbitrary code execution under the user account invoking npx. The package could access user-readable files, environment variables, tokens exposed to the process, or local travel-profile data, and could modify files or initiate n ...[truncated 221 chars]

Remediation
View remediation

Remediation Suggestions

  • Establish and document one authoritative, verified package identity.
  • Remove all references to any incorrect or unexplained package namespace.
  • Replace @latest with an audited exact version.
  • Verify the package publisher, source repository, signatures, and integrity digest.
  • Avoid runtime downloads through npx; install a locked local dependency through a reviewed deployment process.
  • Run the CLI in a sandbox with narrowly restricted filesystem and network permissions.
  • Add automated documentation checks that reject inconsistent dependency names and mutable version tags.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (38)

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The referenced documentation describes a general travel semantic search capability (hotels, attractions, flights, trains) that does not match the declared visa-planning skill purpose. This mismatch can cause the agent to invoke or expose unrelated capabilities, expanding scope and creating a confused-deputy risk where user data or actions are routed through an unintended tool.

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 users to set NODE_TLS_REJECT_UNAUTHORIZED=0, which disables TLS certificate validation for the command. This enables man-in-the-middle interception or tampering of network traffic and normalizes an unsafe workaround far outside the reasonable scope of a visa assistant.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

Advising users to disable TLS certificate verification without strong warnings or compensating controls is an unsafe security practice. It can expose sensitive queries and returned data to interception or manipulation, and it trains operators to ignore certificate failures rather than resolve them securely.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The referenced file documents flight-search functionality, which is materially unrelated to the skill’s stated purpose of visa timeline planning and reminders. This kind of capability mismatch can cause the agent to invoke the wrong tool, mishandle user intent, or expose unrelated travel-booking behaviors/data paths in a context where users expect visa guidance.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The reference file documents hotel-search behavior that is unrelated to the stated visa-planning skill, creating a scope mismatch between declared purpose and available capability. In an agent setting, unrelated references can cause unintended tool use, data handling beyond user expectations, or capability smuggling that weakens trust and policy boundaries.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The referenced file is clearly for Marriott hotel search while the skill metadata describes a visa timeline planner. This kind of capability mismatch can cause the agent to invoke an unrelated tool, leading to incorrect task execution, user confusion, and possible unintended processing of travel-related personal data in the wrong context.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The reference file documents a completely different capability (Marriott package search) than the declared visa timeline planning skill. This mismatch can misroute user requests, cause the agent to invoke unrelated actions, and create a confused-deputy condition where users receive irrelevant or unsafe behavior under the guise of a visa assistant. In this context, the discrepancy is especially dangerous because the skill is supposed to provide travel-document planning, but its embedded reference describes commerce/hotel-package search instead.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The referenced documentation describes a train-search capability, while the skill metadata claims the skill plans visa timelines and reminders. This capability mismatch is dangerous because an agent may invoke the wrong tool or expose unrelated travel-search behavior when handling visa-planning requests, causing unsafe routing, incorrect outputs, and possible policy/control bypass if downstream permissions differ by skill purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The document defines persistent cross-platform storage of broad travel-profile data that goes beyond what a visa timeline assistant needs to answer one-off visa planning questions. This creates unnecessary collection and retention of personal data, increasing privacy risk and expanding the skill's effective scope without a clear purpose limitation.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

Using Qoder Memory to create or update a long-term user profile introduces durable profiling across sessions and potentially across tasks, despite the skill being described as a visa timeline assistant. This broadens data usage beyond the stated function and may preserve sensitive preference and family-travel details without sufficient necessity or privacy controls.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The local-file section prescribes maintaining a persistent travel profile with history, family details, preferences, and other personal data in a reusable cross-session file. For a visa-planning skill, this is unjustified data retention that can expose sensitive personal information if the host environment, filesystem, or other tools/processes access the file.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The activation description includes broad phrases such as "签证时间", "签证材料", and especially "办签证", which can occur in many ordinary travel conversations without clearly indicating the user wants this specific skill invoked. The file also does not provide exclusion conditions or negative examples to narrow when the skill should not activate.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

The manifest description and trigger phrases are entirely in Chinese, presenting the skill as operating in a specific language without any indication that users may choose another language. Under the policy, forcing a specific language without user opt-in is a natural-language policy concern unless the locale limitation is explicitly justified.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill instructs the agent to read historical user preferences and profile data even though visa timeline generation only requires current trip and identity-related inputs. Accessing stored profile data without a strong necessity increases privacy exposure and risks collecting or using personal information beyond the minimum needed for the task.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill describes reading historical preferences and a local profile file without an explicit user-facing privacy warning or consent step. In context, this makes the data access more dangerous because the capability is hidden behind startup behavior rather than being transparently requested when needed.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Using a local file fallback such as ~/.flyai/user-profile.md gives this skill filesystem access unrelated to its core visa-planning purpose. That increases the blast radius from simple preference recall to local data exposure, especially if the file contains broader personal information or if path handling is later generalized.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill’s stated purpose is visa timeline planning, but it expands into visa agency search and generation of flight/hotel booking artifacts. This broadens the operational scope and can push the agent to handle adjacent actions involving third-party commerce and potentially sensitive travel details without a clearly bounded need for the core task.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

Persisting new user preferences extends the skill from one-time planning into ongoing profile management, which is broader than the described timeline/reminder function. This creates unnecessary retention of user data and raises privacy and consent concerns if the saved data is not tightly scoped.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The workflow mandates Chinese-language questions, options, and output templates throughout the file, which effectively forces a specific language. Under the stated policy, locale or language constraints should either be opt-in or clearly documented as justified for a region-specific tool.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The workflow requires installing or upgrading a global CLI before performing a simple visa-planning task, and even suggests using sudo if permissions fail. This creates an unnecessary supply-chain and privilege-escalation path unrelated to the stated skill purpose, increasing the chance of system modification or compromise if the package, registry path, or execution environment is unsafe.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
96% confidence
Finding

The workflow explicitly recommends running npm installation with sudo, which elevates package installation to root privileges. If the package or its dependency chain is malicious or compromised, this can lead to full-system compromise rather than a contained user-level impact.

Content

Scanner excerpt · reference/core-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 is an unsafe default because it disables certificate verification globally for that process. In context, this is especially dangerous because the skill is a travel-planning assistant, so users would not reasonably expect to weaken transport security to complete the task.

Content

Scanner excerpt · reference/core-workflow.md (reported line 112)May include surrounding context.

md
> **⚠️ SSL 证书问题**:如果通过终端执行 flyai CLI 命令时遇到 `UNABLE_TO_GET_ISSUER_CERT_LOCALLY` 错误,需要在命令前加上环境变量:
> ```bash
> NODE_TLS_REJECT_UNAUTHORIZED=0 flyai keyword-search --query "{目的地国家} 签证"
> ```

从查询结果中提取:

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

manifest 说明该技能用于倒推签证办理时间线、生成材料清单和每步截止日期;而 L233-L243 进一步加入 search-flight 和 search-hotel 查询,为机票酒店预订单提供帮助。这超出了签证进度规划的核心描述范围,构成描述与实际工作流能力的不一致。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The examples and output are entirely localized in Chinese, including city names, currency (CNY/¥), weekday names, and flight descriptors, with no indication that users can choose another language or locale. This suggests a locale-specific behavior that may violate policy if the skill is intended for general use without explicit opt-in or justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The documentation specifies bed-type values only in Chinese and all invocation examples use Chinese destination and hotel names. This effectively constrains users to a specific language/locale without any opt-in or note that the skill is region-specific.

Content

No source excerpt is available for this finding.