Back to skill

Security audit

通联手续费发票申请

Security checks for vulnerabilities and agentic risk

Overview

This skill automates a disclosed invoice-application workflow, but users should verify recipients, tax details, and any remembered customer preferences before use.

Before installing, confirm that the missing template files are supplied elsewhere, require trusted invoice/tax details instead of relying on web lookup, and only allow remembered customer settings if users can review or correct them. Always verify the recipient, attachments, month, amount, and invoice type before approving the email send.

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
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 Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (5)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases are broad enough that common billing-related messages or attached statements could activate a workflow that reads invoices, extracts financial data, and prepares outbound email. In a skill that processes sensitive financial records and can send email on the user's behalf, ambiguous invocation materially increases the chance of unintended execution and data handling.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly instructs the agent to obtain enterprise invoicing data from the internet, which expands data collection beyond user-provided billing artifacts and approved internal sources. This creates a risk of using inaccurate or tampered third-party information in financial documents and may cause unauthorized external data access during a workflow that handles sensitive billing information.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill instructs persistent storage of customer-specific invoicing preferences and recipient email addresses for future reuse without defining retention limits, access controls, or user consent. Persisting financial preferences and contact data creates unnecessary long-term privacy exposure and raises the risk of stale or misapplied settings causing data to be sent to the wrong party.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill directs the agent to send billing and invoice attachments to external recipients on the user's behalf without an explicit privacy warning or clear consent language about external transmission. Because the workflow includes customer billing data, merchant identifiers, tax details, and recipient email addresses, omission of data-sharing safeguards increases the risk of accidental disclosure or unauthorized transfer.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

Lines L142-L146 and the worked example at L154 instruct using MMDD in the email body, but L170 says the body date format must uniformly be YYMM instead of MMDD. This is an active contradiction in the skill documentation about what the generated email should contain.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.