Back to skill

Security audit

捷帮SEO分析

Security checks for vulnerabilities and agentic risk

Overview

This SEO skill does what it claims, but it forwards full user-provided URLs to an external service using an admin-style credential without clear disclosure or input safeguards.

Review before installing. Use this only with public webpages whose full URL, including path and query string, can safely be shared with www.jiebang.site. Do not submit private dashboards, signed links, reset links, authenticated pages, internal hostnames, or URLs containing tokens. The publisher should document the external processing and replace the admin-style credential with a narrower API token.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

other

Warning
Location
scripts/main.py:14
Finding

Undisclosed Transmission of User-Supplied URLs to a Third-Party Service

Content
View full analysis

Vulnerability Details

File Location: scripts/main.py, lines 14 and 41–86
Vulnerability Type: Undisclosed Third-Party Data Transmission
Risk Level: Medium

Vulnerable Code

python
BASE_URL = "https://www.jiebang.site"

Metadata extraction transmits the supplied URL to the external service:

python
api_url = f"{BASE_URL}/api/meta-extract"

try:
    params = {"url": url}
    response = requests.get(
        api_url,
        headers=get_headers(),
        params=params,
        timeout=30
    )

SEO checking performs the same transmission:

python
api_url = f"{BASE_URL}/api/seo-check"

try:
    params = {"url": url}
    response = requests.get(
        api_url,
        headers=get_headers(),
        params=params,
        timeout=30
    )

If either GET request fails with an HTTP error, the URL is transmitted again in a POST body:

python
response = requests.post(
    api_url,
    headers=get_headers(),
    json={"url": url},
    timeout=30
)

Technical Analysis

Both supported operations send the complete user-controlled URL to https://www.jiebang.site. The Skill documentation describes metadata extraction and SEO analysis but does not disclose that the URL is processed by an external third party.

A URL can contain information beyond a public hostname, including private hostnames, sensitive paths, user identifiers, signed query parameters, password-reset tokens, presigned object-storage credentials, API keys, or analytics identifiers. Because the input is forwarded without redaction or validation, all such components can be received and logged by the external service and its infrastructure.

HTTPS protects the request while it is in transit, but it does not prevent the destination service from accessing, retaining, or processing the URL. The code also retries an unsuccessful GET request as a POST request, potentially trans ...[truncated 1843 chars]

Remediation
View remediation

Remediation Suggestions

  1. Clearly disclose in SKILL.md that submitted URLs are sent to www.jiebang.site for remote processing, including what data is transmitted and how it may be retained.
  2. Require explicit user consent before transmitting a URL to the third-party service.
  3. Parse URLs locally and reject embedded credentials such as https://user:password@example.com/.
  4. Remove query parameters and fragments by default. If query parameters are required, use an allowlist and redact common secret-bearing fields such as token, key, signature, auth, code, and session.
  5. Warn users not to submit private, authenticated, signed, or password-reset URLs.
  6. Validate the scheme and accept only http and https URLs.
  7. Resolve and block loopback, link-local, private, reserved, and internal destinations unless internal scanning is explicitly intended and safely controlled.
  8. Avoid automatic retransmission of sensitive input. If a GET request fails, obtain consent before retrying through another method or ensure the retry cannot create additional exposure.
  9. Prefer local metadata parsing where feasible so that user URLs do not need to be disclosed to an external service.
  10. Establish and document appropriate retention, access-control, and log-redaction policies for the receiving API.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (6)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding

The skill appears to perform network fetching of arbitrary user-supplied URLs, but it does not declare any explicit tool scope or permission boundaries. That creates a real security gap because reviewers and runtime policy engines cannot clearly constrain or audit whether network and environment access are required, increasing the risk of SSRF-style misuse or unintended data exposure if the implementation fetches attacker-controlled targets.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The invocation description uses broad trigger phrases like SEO analysis, metadata extraction, and optimization suggestions without clear boundaries on what URLs or tasks are in scope. In practice, overly broad routing can cause the skill to activate on loosely related requests and fetch arbitrary URLs, which increases exposure to untrusted network targets and makes misuse easier.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The markdown does not warn users that the skill will retrieve and inspect a user-provided URL over the network. This matters because users may unknowingly cause requests to third-party or sensitive endpoints, and the lack of disclosure reduces informed consent while obscuring SSRF/privacy implications of supplying a URL.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill uses an environment-scoped admin credential and attaches it to every request for routine SEO and metadata operations. For a tool that only submits a target URL to backend analysis endpoints, using an admin-level key violates least-privilege and increases blast radius if the skill, logs, downstream service, or request path are compromised.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill forwards user-supplied URLs to an external service together with an admin credential, without any visible disclosure, consent, or validation. This creates a privacy and trust boundary issue, and if attackers can influence the URL input they may cause sensitive internal or user-provided URLs to be transmitted to a third party under elevated authentication context.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The module description and CLI help text are presented in Chinese only, which can amount to forcing a specific language for users without offering alternatives. The file does not document that the skill is intentionally region-specific or provide any language selection mechanism.

Content

No source excerpt is available for this finding.