ClawHub
Back to skill

Security audit

Save Token

Security checks for vulnerabilities and agentic risk

Overview

This skill is a documentation-only token-saving guide with broad activation wording but no executable code, persistence, credential access, or hidden data flow.

Installers should understand that this skill may encourage agents to summarize or omit repeated context to save tokens. That can be useful for cost and speed, but users should ask for full context to be preserved when precision, auditability, safety, or legal/financial/medical details matter.

A.I.G

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

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (6)

Natural-Language Policy Violations

Medium
Confidence
88% confidence
Finding
The entire skill documentation is presented only in Chinese and does not indicate any user language choice or opt-in. This can violate language or locale policy when users are not explicitly given the option to use another language.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The listed triggers include common terms like "成本" and "省钱" as well as broad contextual conditions such as long conversations or repeated file reads, without clear boundaries or exclusion conditions. This makes it unclear when the skill should activate versus remain inactive, increasing the chance of accidental invocation in ordinary conversations.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The activation conditions are expansive, including routine states like long conversations, multiple file reads, or complex tasks. Without stricter boundaries, the skill may be invoked too often and encourage unnecessary compression of context, which can degrade task fidelity or omit important details.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The trigger keyword list is broad and includes generic terms like 'token', 'cost', 'context', and optimization-related words in two languages. This can cause the skill to activate in situations outside its intended scope, leading the agent to apply summarization or context-reduction guidance when precision or completeness would be preferable.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
| Cannot Do | Reason | 不能做 | 原因 |
|-----------|--------|--------|------|
| Directly modify context | Skill is guidance, not execution | 直接修改上下文 | Skill 是指导，不是执行 |
| Automatically delete content | Agent must decide what to keep | 自动删除内容 | Agent 必须决定保留什么 |
| Guarantee exact savings | Depends on implementation | 保证精确节省量 | 取决于实现方式 |

## ✅ What This Skill Does | 此 Skill 能做的事
Confidence
80% confidence
Finding
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Natural-Language Policy Violations

Low
Confidence
81% confidence
Finding
The skill consistently presents bilingual English and Chinese content, but it does not state that the user can choose their preferred language or that the bilingual format is required for a specific audience. This may conflict with a language/locale policy that avoids imposing a language choice without opt-in.

Static analysis

No suspicious patterns detected.