Back to skill

Security audit

Event Travel Planner

Security checks for vulnerabilities and agentic risk

Overview

This event-planning skill is mostly purpose-aligned, but its setup and login flow grant risky system and account access that users should review before installing.

Install only if you are comfortable with non-interactive dependency installation, global or persistent CLI changes, third-party travel/search queries, and Xiaohongshu account-cookie use. Prefer manual setup with pinned versions, review the setup script first, use QR login instead of browser cookie extraction when possible, and validate Fly.ai booking links before presenting or clicking them.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Findings (5)

T03 · Remote Payload Retrieval and Execution

Error
Location
scripts/setup.sh:100
Finding

Unverified Remote Installer Is Piped Directly to a Shell

Content
View full analysis
/dev/null | sh 2>/dev/null; then ``` The error guidance at lines 107 and 111 also recommends executing: ```bash curl -LsSf https://astral.sh/uv/install.sh | sh ``` ### Technical Analysis The setup script downloads mutable content from an external URL and pipes it directly to `sh`. It does not pin an installer version, verify a cryptographic digest or signature, save the script for inspection, or obtain explicit approval immediately before execution. Although `astral.sh` is associated with the declared `uv` dependency, reliance on HTTPS and domain reputation alone does not establish the integrity of mutable executable content. A compromised origin, CDN, DNS path, certificate trust chain, or upstream deployment could alter the effective payload after this Skill has been reviewed. Redirecting standard error to `/dev/null` also suppresses information that could reveal unexpected installer behavior or failures. ### Attack Path 1. The user or Agent runs `scripts/setup.sh`. 2. The script determines that neither `uv` nor `pipx` is installed. 3. The script requests `https://astral.sh/uv/install.sh`. 4. The upstream response is compromised or changed maliciously. 5. The response is passed directly to `sh` without integrity verification. 6. The payload executes immediately with all permissions of the user running the setup script. ### Impact Assessment A malicious installer could execute arbitrary commands under the Agent user's account. Depending on that user's permissions, it could read or modify project files, shell configuration, credentials, browser data, API keys, and user-owned files. If the setup script is run by a privileged account, the impact could extend to system-wide compromise. ]]>
Remediation
View remediation

T08 · Insecure Dependencies

Error
Location
scripts/setup.sh:118
Finding

Automatic Installation of Unpinned Third-Party CLI Packages

Content
View full analysis
/dev/null; then XHS_INSTALLED="yes" else if uv tool upgrade xiaohongshu-cli 2>/dev/null; then XHS_INSTALLED="yes" fi fi ``` ```bash if pipx install xiaohongshu-cli 2>/dev/null; then XHS_INSTALLED="yes" else if pipx upgrade xiaohongshu-cli 2>/dev/null; then XHS_INSTALLED="yes" fi fi ``` ```bash if npm install -g @fly-ai/flyai-cli 2>/dev/null; then FLYAI_INSTALLED="yes" else FLYAI_INSTALLED="no" fi ``` ### Technical Analysis The setup process resolves package names to their latest available registry versions without pinning exact versions or verifying package integrity against an independently trusted value. It can also automatically upgrade `xiaohongshu-cli`, meaning reviewed behavior may be replaced by a later release. The npm package is installed globally. This exceeds the isolation needed for a single Skill and expands the package's reach into the user's global command environment. Package installation may execute package-controlled installation hooks with the permissions of the invoking user. This creates a supply-chain risk even if the package names currently resolve to legitimate projects. Registry account compromise, malicious package updates, dependency confusion within transitive dependencies, or compromised maintainer credentials could introduce executable code after the Skill audit. ### Attack Path 1. One of the required CLIs is absent. 2. The Agent runs the automatic setup script. 3. The package manager resolves an unpinned latest release and its transitive dependencies. 4. A compromised release or dependency is downloaded from the package registry. 5. Package inst ...[truncated 589 chars]
Remediation
View remediation

T01 · Skill Instruction Hijacking

Warning
Location
scripts/xhs_filter.py:80
Finding

Untrusted Social-Media Content Is Inserted Directly into Agent Context

Content
View full analysis
Remediation
View remediation

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md:163
Finding

Skill Instructions Mandate Commercial Links and Brand Promotion

Content
View full analysis
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Error
Location
references/xhs_cli.md:31
Finding

Unpinned Third-Party CLI Is Directed to Extract Browser Session Cookies

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (19)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

声明描述的是一个面向线下活动出行的综合助手,核心能力应包括活动相关信息检索、小红书攻略整合、以及飞猪上的交通住宿预订与行程安排。但提供的代码只处理小红书 CLI 返回的 JSON:search 模式提取标题、发布时间和互动数据;read 模式提取标题、正文与时间;comments 模式提取评论内容与点赞数。其主要用途是将小红书数据压缩成更易供 AI 阅读的文本格式。代码没有显示任何飞猪接口、预订逻辑、行程规划算法、活动场次处理、场地周边联动、触发词实现或多工具编排。因此,代码行为与声明的主要目的存在明显且实质性的不匹配。

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill instructs the agent to execute multiple shell commands (date, xhs, flyai, pipes, grep) but does not declare any explicit tool scope such as allowed tools or permissions. This creates an authorization gap where a reviewer or runtime cannot easily constrain execution, increasing the chance of unintended command execution or privilege overreach if the skill is invoked in a broader environment.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases include broad everyday expressions such as '参加活动' and '看演出', which can cause the skill to activate in unrelated conversations. In a skill that performs shell commands and external searches, overbroad activation increases the risk of unexpected tool use, data exposure to third-party services, or user confusion without clear intent.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
77% confidence
Finding

文件整体以中文定义技能用途和触发词,且未说明可根据用户偏好切换语言,也未声明这是仅面向中文用户或中国区场景的限定工具。按规则,这属于未提供用户选择的语言/locale 约束。

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill mandates generating an HTML file for the user but does not disclose that it will create or write a local file, nor does it describe where the file will be stored. Silent file creation is risky because it can surprise users, overwrite existing files, or introduce unsafe rendered content if external data is embedded into HTML without sanitization.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The documentation requires the assistant to include fixed Chinese-language branding text ("基于 fly.ai 实时搜索结果") with no user-language choice. This reduces user autonomy and can cause deceptive or low-quality output in multilingual contexts, especially if the user requested another language or a neutral response. In this event-planning skill, the issue is less dangerous than code-execution or data-exfiltration flaws, but it still introduces hard-coded output manipulation that may conflict with user intent and platform policy.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The guide directs the agent or user to run an automatic setup script that may install software and inspect authentication state, but it does not require explicit confirmation, preview the changes, or constrain what the script may do. In an agent skill context, this increases the risk of unintended system modification, supply-chain exposure, or execution of unsafe repository code during setup.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The document instructs the agent to extract browser-stored Xiaohongshu cookies via xhs login and browser auto-detection, but it does not clearly require explicit user consent or warn that browser credentials will be accessed. In an agent context, this can normalize silent access to account-linked session material and lead to privacy violations or misuse of authenticated sessions if the agent performs login flows without informed approval.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The reference describes external commands that send search terms, note URLs, comments, and authenticated session context to Xiaohongshu, but it does not clearly disclose these privacy implications to the user. Because this skill is designed for event planning and may process sensitive travel or attendance interests, users may unknowingly expose behavioral data, account-linked activity, and shared URLs to a third-party service.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The script's comments, banners, prompts, guidance, and final status messages are all written in Chinese, with no opt-in or language-selection mechanism. This imposes a fixed language/locale on all users and agents rather than allowing user preference.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The script is explicitly designed for full automation and proceeds to install tools globally and fetch dependencies without any confirmation gate, dry-run mode, or safety prompt. In an agent-executed context this can modify the host environment unexpectedly, increase supply-chain risk, and violate least surprise for users who may not realize package installation will occur.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This code file prints full post content and comment metadata derived from xhs-cli JSON, including nicknames and free-form comment text. Although the script's docstring explains its purpose, it does not disclose that personal or sensitive user-generated content will be surfaced for AI consumption, which is a privacy-relevant behavior under the missing user warnings rule.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

A language-specific instruction set can violate language/locale policy when it implicitly forces one language without user opt-in. This file provides all operational guidance only in Chinese and does not indicate that Chinese is optional, user-selected, or required for a justified region-specific reason.

Content

No source excerpt is available for this finding.

External Script Fetching

Low
Category
Supply Chain
Confidence
98% confidence
Finding

The script executes a remote installer directly via curl | sh, which gives code from an external server immediate shell execution on the host without integrity verification. In an automated agent context, this is especially dangerous because a compromised upstream, MITM, or unexpected content change can lead to arbitrary code execution during setup.

Content

Scanner excerpt · scripts/setup.sh (reported line 100)May include surrounding context.

sh
HAS_PIPX="yes"
    else
      warn "未找到 uv 或 pipx,将尝试自动安装 uv..."
      if curl -LsSf https://astral.sh/uv/install.sh 2>/dev/null | sh 2>/dev/null; then
        export PATH="$HOME/.local/bin:$HOME/.cargo/bin:$PATH"
        if command -v uv &>/dev/null; then
          ok "uv 已自动安装:$(uv --version 2>/dev/null | head -1)"

External Script Fetching

Low
Category
Supply Chain
Confidence
15% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · scripts/setup.sh (reported line 107)May include surrounding context.

sh
HAS_UV="yes"
        else
          err "uv 自动安装后仍不可用"
          ERRORS="${ERRORS}\n- [xhs-cli] uv 安装失败,请手动执行:curl -LsSf https://astral.sh/uv/install.sh | sh"
        fi
      else
        err "uv 自动安装失败"

External Script Fetching

Low
Category
Supply Chain
Confidence
15% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · scripts/setup.sh (reported line 111)May include surrounding context.

sh
HAS_UV="yes"
        else
          err "uv 自动安装后仍不可用"
          ERRORS="${ERRORS}\n- [xhs-cli] uv 安装失败,请手动执行:curl -LsSf https://astral.sh/uv/install.sh | sh"
        fi
      else
        err "uv 自动安装失败"

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The script inspects two credential sources for FlyAI API keys: ~/.flyai/config.json and the FLYAI_API_KEY environment variable. Even though it only checks presence rather than exfiltrating the secret, reading credential locations is a sensitive capability and is not clearly necessary for a setup script whose core purpose is event-planning assistance; it also occurs without an explicit user-facing disclosure before inspection.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The script inspects sensitive credential sources without a clear prior notice that it will read a config file and environment variable. While the check appears intended to improve UX, undisclosed secret-source inspection is risky in agent settings because users may not expect setup code to probe authentication material.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

Natural-language strings throughout the file, including the module description, usage text, and output labels, are exclusively in Chinese. This forces a specific language experience without any opt-in or explanation that the tool is intentionally limited to a Chinese-speaking or region-specific context.

Content

No source excerpt is available for this finding.