Back to skill

Security audit

猫棒状态

Security checks for vulnerabilities and agentic risk

Overview

The skill is a clear GPON status helper, but it tells an agent to log in as root over unencrypted HTTP with a blank password and reuse the session cookie without safeguards.

Install only if you understand that the skill will access a local device admin page as root over HTTP. Use it only on a trusted, isolated local network, confirm the IP belongs to your intended GPON/OpenWrt device, avoid logging cookies or credentials, and prefer HTTPS or a read-only monitoring account if the device supports them.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:16
Finding

Blank Root Authentication and Session Transported over Plaintext HTTP

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 16–17
Vulnerability Type: Plaintext privileged authentication and insecure session transport
Risk Level: High

Vulnerable code snippet:

text
POST http://192.168.11.6/cgi-bin/luci
username=root&password=

The subsequent instruction reuses the resulting session cookie to access the same web interface.

Technical Analysis

The Skill instructs the agent to authenticate to a device-management interface as the privileged root user with an empty password. The request is sent through unencrypted HTTP, and the resulting session cookie is reused for subsequent access.

HTTP provides neither transport confidentiality nor endpoint integrity. An attacker with visibility or control over the local network path can inspect the credentials and session cookie, modify the authentication response, inject content into later HTML responses, or replay a captured cookie. A blank password also removes a primary authentication factor and makes the administrative interface accessible to any party that can reach it if no additional safeguards are present.

The target address is fixed rather than explicitly confirmed or securely configured. In a different network environment, 192.168.11.6 could belong to an unintended or attacker-controlled host, which could collect the login request and return forged status data.

Attack Path

  1. A user invokes the Skill while connected to a network containing the configured address.
  2. The agent sends username=root&password= to http://192.168.11.6/cgi-bin/luci.
  3. An attacker on the same network, a compromised gateway, or a malicious device assigned that address observes or modifies the plaintext exchange.
  4. The attacker captures the administrative session cookie or impersonates the management endpoint.
  5. The attacker replays the cookie against the device, where permitted, or returns manipulated HTML that causes ...[truncated 1081 chars]
Remediation
View remediation

Remediation Suggestions

  1. Enable HTTPS on the management interface and require certificate verification. Do not silently fall back to HTTP.
  2. Replace the blank root password with a strong, unique credential. Obtain it at runtime from an approved secret store or protected environment variable rather than embedding it in Skill instructions.
  3. Create a dedicated read-only monitoring account with access only to the status information required by the Skill. Avoid using root.
  4. Make the target address an explicit user configuration and require confirmation that the endpoint is the intended device. Where possible, pin an expected certificate or device identity.
  5. Protect session cookies with Secure, HttpOnly, and appropriate SameSite attributes, rotate the session after authentication, impose a short expiration, and invalidate it after the status query.
  6. Prevent credentials and cookie values from appearing in logs, command histories, reports, or error output.
  7. If the device cannot support secure transport, access it through a trusted isolated management network or a verified encrypted tunnel and clearly warn users that direct HTTP operation is unsafe.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill instructs an agent to authenticate to a device web interface over plain HTTP and reuse the resulting session cookie to scrape operational data, but provides no warning, consent gate, or credential/session safeguards. This can expose credentials or session tokens on the local network, encourage automatic access to sensitive infrastructure, and normalize scraping device admin pages without explicit user authorization.

Content

No source excerpt is available for this finding.