Back to skill

Security audit

Code Pro

Security checks for vulnerabilities and agentic risk

Overview

This is a plain workflow guide for code analysis, generation, review, and debugging, with no executable payload or hidden data access.

Install this if you want a structured coding workflow. Be aware that broad phrases like debug, implement, or full review may activate more detailed analysis than expected, so specify the files, directories, and stopping point when working in sensitive repositories.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The debug-mode triggers include broad phrases like '调试', 'debug', '修复问题', and especially '接着跑', which are common in normal conversations and can activate iterative behavior unintentionally. In this skill, that matters because the mode explicitly encourages repeated attempts and continuation behavior, which can cause the agent to keep acting on tasks beyond the user's precise scope or after an unsafe intermediate step.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The long-context mode is activated by vague phrases such as '大项目', '完整分析', and '全面审查', all of which can appear in ordinary requests. In a code-analysis skill, accidental activation of repository-wide analysis can broaden the model's operational scope, increase exposure to unrelated sensitive content, and make it easier for adversarial user content to route the agent into more expansive behavior than intended.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The quick-command table maps very common user intents such as '写新代码', '修bug', and '全面评估' to this skill without clear activation boundaries. Because these phrases overlap with ordinary coding requests, the skill can be invoked unexpectedly and steer the agent into a more powerful workflow than the user explicitly requested, increasing the chance of unintended behavior or policy bypass through prompt routing confusion.

Content

No source excerpt is available for this finding.