Back to skill

Security audit

审计调整分录核对(免费版)

Security checks for vulnerabilities and agentic risk

Overview

The local checker appears self-contained, but the skill also embeds an external paid upgrade/install flow with plain HTTP purchase routing that deserves review before installation.

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

Warning
Location
SKILL.md:125
Finding
Unauthenticated HTTP endpoints control paid-product metadata and checkout navigation## Vulnerability Details **File Location**: `SKILL.md`, lines 125–126 **Vulnerability Type**: Unauthenticated purchase-routing channel **Risk Level**: Medium **Vulnerable snippet**: ```text http://110.40.221.75/buy/audit-adjustment-check http://110.40.221.75/buy/audit-adjustment-check?format=json http://110.40.221.75/buy/audit-adjustment-check?go=1 ``` ### Technical Analysis The skill documents an unauthenticated HTTP endpoint as a permanent purchase address. Its query variants supply the current product identifier as JSON or issue an HTTP redirect to the purported current SkillPay purchase destination. Because plain HTTP provides neither server authentication nor transport integrity, an on-path attacker can modify the JSON response, replace the product metadata, or substitute the redirect target. The skill text instructs the agent to verify merchant and product information, but this is not an enforced technical control and does not authenticate the HTTP response itself. The local audit engine does not contact these endpoints. The vulnerable path is triggered only when a user chooses the optional paid version and the agent or user follows the documented HTTP purchase workflow. ### Attack Path 1. A user selects the optional paid version. 2. The agent or user accesses the documented HTTP endpoint to obtain the current product identifier or redirect destination. 3. An attacker with an on-path network position intercepts the unencrypted request. 4. The attacker replaces the JSON metadata or HTTP redirect with attacker-controlled purchase information. 5. The agent presents, or the user follows, the substituted checkout destination. 6. If the documented manual merchant verification is missed or bypassed, the user may disclose payment-related information to a phishing destination or purchase an unintended product. ### Impact Assessment Successful exploitation can redirect the paid upgrade workflow to an attacker-con ...[truncated 393 chars]
Remediation
## Remediation Suggestions - Remove the plain HTTP endpoints and expose purchase metadata and redirects exclusively through HTTPS with valid certificate verification. - Prefer an authenticated official platform API rather than a mutable intermediary endpoint. - Enforce an allowlist for the final checkout scheme and hostname after every redirect. - Programmatically verify the expected merchant ID, product ID, order amount, and checkout origin before presenting a payment link. - Reject redirects to raw IP addresses, non-HTTPS destinations, unexpected hosts, or mismatched merchant metadata. - Keep payment confirmation under explicit user control and fail closed if any identity or amount check cannot be completed.

Static analysis

No suspicious patterns detected.