Back to skill

Security audit

xingqiaoskill

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent remote messaging client, but it sends account credentials and user content over plaintext HTTP and stores reusable tokens in a local plaintext file.

Review this skill carefully before installing. Avoid sending sensitive content through it unless the service moves to HTTPS and improves credential storage. If used, treat config.json as a secret, do not commit or share it, and rotate/delete the token if the file or network traffic may have been exposed.

Vulnerability Patterns
  • 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
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (3)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/install.py:14
Finding

Authentication Credentials and User Content Transmitted Over Plaintext HTTP

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/install.py:52
Finding

Reusable Authentication Secrets Stored in an Unprotected Plaintext Configuration File

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
SKILL.md:10
Finding

Unpinned Third-Party Dependency Produces Non-Reproducible Installations

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (22)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

This finding identifies a stronger behavior mismatch: the skill claims natural-language command handling and trigger constraints, but the actual behavior reportedly centers on remote registration/login, JWT retrieval, and writing credentials to config.json. Such divergence is a real security issue because it conceals sensitive authentication flows and persistence behavior behind an innocuous description.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

This finding identifies a stronger behavior mismatch: the skill claims natural-language command handling and trigger constraints, but the actual behavior reportedly centers on remote registration/login, JWT retrieval, and writing credentials to config.json. Such divergence is a real security issue because it conceals sensitive authentication flows and persistence behavior behind an innocuous description.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The installer uses a hard-coded HTTP endpoint and sends the generated skill token to /api/auth/login as a URL query parameter, then receives a JWT over plaintext transport. This exposes both the token identifier and returned authentication material to interception or modification by any attacker on the network path, and query-string credentials may also be logged by proxies and servers. In this skill context, the installer bootstraps the account used by the messaging/information platform, so compromise at install time can lead to persistent account takeover and unauthorized message access or impersonation.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
94% confidence
Finding

The skill declares installation metadata and describes behaviors that require network access and local file writes, but it does not declare an explicit tool/permission scope. This weakens user consent and platform enforcement because the skill can register remote accounts and persist tokens/config locally without a clearly bounded permission model.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The first-use flow automatically creates a remote account, generates a token, and stores configuration locally, but the skill does not present this as a prominent privacy/security warning before use. Users may unknowingly create an external identity and leave credentials on disk, which increases privacy and account-abuse risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill sends messages, subscriptions, replies, and related user content to a remote service at an HTTP endpoint, yet the documentation lacks a strong privacy/security warning. Because the configured API base uses plain HTTP, transmitted content and tokens may be exposed to interception or tampering, making the omission more dangerous in this context.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The file's natural-language interface, help text, and command parsing are entirely Chinese-centric, including the docstring and usage examples, with no indication that users may opt into another language. This can violate language or locale policy when a skill mandates a specific language without documented justification or user choice.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The initialization flow stores the JWT and long-lived token_id in a local config file without warning the user or applying file-permission hardening. If the local filesystem is accessible to another user, process, backup system, or malware, these credentials can be stolen and used to impersonate the skill account and access or post data remotely.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · scripts/cli.py (reported line 149)May include surrounding context.

python
data = {"content": content}
    
    try:
        response = requests.post(url, json=data, headers=headers)
        if response.status_code == 200:
            result = response.json()
            print(f"[OK] 发送成功!")

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · scripts/cli.py (reported line 220)May include surrounding context.

python
data = {"content": content}
    
    try:
        response = requests.post(url, json=data, headers=headers)
        if response.status_code == 200:
            result = response.json()
            print(f"[OK] 发送成功!")

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · scripts/cli.py (reported line 253)May include surrounding context.

python
data = {"content": content}
    
    try:
        response = requests.post(url, json=data, headers=headers)
        if response.status_code == 200:
            result = response.json()
            print(f"[OK] 发送成功!")

Tainted flow: 'data' from requests.post (line 33, network input) → requests.post (network output)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/cli.py (reported line 149)May include surrounding context.

python
data = {"content": content}
    
    try:
        response = requests.post(url, json=data, headers=headers)
        if response.status_code == 200:
            result = response.json()
            print(f"[OK] 发送成功!")

Tainted flow: 'data' from requests.post (line 33, network input) → requests.post (network output)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/cli.py (reported line 220)May include surrounding context.

python
data = {"content": content}
    
    try:
        response = requests.post(url, json=data, headers=headers)
        if response.status_code == 200:
            result = response.json()
            print(f"[OK] 发送成功!")

Tainted flow: 'data' from requests.post (line 33, network input) → requests.post (network output)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/cli.py (reported line 253)May include surrounding context.

python
data = {"content": content}
    
    try:
        response = requests.post(url, json=data, headers=headers)
        if response.status_code == 200:
            result = response.json()
            print(f"[OK] 发送成功!")

Tainted flow: 'data' from requests.post (line 33, network input) → requests.post (network output)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/cli.py (reported line 319)May include surrounding context.

python
data = {"content": content}
    
    try:
        response = requests.post(url, json=data, headers=headers)
        if response.status_code == 200:
            result = response.json()
            print(f"[OK] 发送成功!")

Tainted flow: 'data' from requests.post (line 33, network input) → requests.post (network output)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/cli.py (reported line 394)May include surrounding context.

python
data = {"content": content}
    
    try:
        response = requests.post(url, json=data, headers=headers)
        if response.status_code == 200:
            result = response.json()
            print(f"[OK] 发送成功!")

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The CLI implements a public Q&A system that is not disclosed in the skill metadata, expanding the skill's capabilities beyond the declared push/pull/subscribe/reply workflow. Hidden or undocumented features increase the chance of unintended data disclosure, broader external posting than users expect, and weakened reviewability of the skill's true behavior.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% confidence
Finding

This line transmits user-provided content to an external public Q&A endpoint that is not declared in the skill metadata. Because the feature is hidden from the manifest, users may unintentionally disclose information publicly when they expect only the documented messaging workflow, making the transmission more dangerous in this skill context.

Content

Scanner excerpt · scripts/cli.py (reported line 319)May include surrounding context.

python
data = {"content": content}
    
    try:
        response = requests.post(url, json=data, headers=headers)
        if response.status_code == 200:
            result = response.json()
            print(f"[OK] 问题发布成功!")

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% confidence
Finding

This call posts user content to an answers endpoint belonging to the undeclared public Q&A feature set. The risk comes from hidden scope expansion: content may be published externally under a workflow not described to users or reviewers, increasing the chance of accidental data leakage.

Content

Scanner excerpt · scripts/cli.py (reported line 394)May include surrounding context.

python
data = {"content": content}
    
    try:
        response = requests.post(url, json=data, headers=headers)
        if response.status_code == 200:
            result = response.json()
            print(f"[OK] 回答成功!")

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The parser recognizes many core actions only through Chinese phrases such as '发送', '订阅', '回复', '重要通知', and '问题列表', while the help output is also Chinese-only. Because the skill does not offer language selection or explicitly justify a locale restriction, it effectively enforces a specific language for use.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The command parser exposes additional commands such as important notifications and public Q&A workflows that are not described in the manifest. This creates a transparency and trust issue because users and reviewers may believe the skill only handles private subscription messaging while it can also interact with other remote features and surface additional data.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The script's docstrings, status messages, and usage instructions are entirely in Chinese, and it presents fixed Chinese command examples to the user. This enforces a specific language experience without any visible opt-in or locale selection, which matches the language-policy violation criteria.

Content

No source excerpt is available for this finding.