Back to skill

Security audit

思考×知识库闭环系统

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed knowledge-base helper, but it can automatically save answers into local wiki files from broad prompts without a clear confirmation step.

Review this skill before installing if you do not want routine analysis or organization requests to modify persistent knowledge-base files. Use it only in workspaces where writing to wiki/, index.md, log.md, and AGENTS.md is acceptable, and prefer asking the agent to preview changes before saving.

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 (5)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The AGENTS.md rules instruct automatic write-back after 'every valuable comprehensive answer' without warning the user that content will be persisted to wiki files. Automatic persistence is dangerous because it can store sensitive, incorrect, or prompt-injected content into a durable knowledge base, expanding the blast radius beyond a single chat.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger phrase for activating the thinking mode is very broad (e.g. ordinary requests like '分析 XXX' or '帮我理清思路'), which can cause the skill to engage unexpectedly during routine conversation. In a skill system, overly generic activation language increases the chance of unintended behavior and can bypass user expectations about when special processing will occur.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The knowledge-base triggers are similarly ambiguous ('写入知识库' or '帮我整理') and may match common user intents that do not imply consent to persistent file operations. This creates a risk that normal assistance requests will invoke state-changing behavior without the user's clear understanding.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The combined-use instructions imply that asking for analysis plus storage will result in automatic wiki modification, but they do not clearly communicate persistence, affected files, or associated risks. This makes it easy for users to trigger lasting side effects without informed consent, especially when paired with the broad trigger language elsewhere in the skill.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The document is primarily written in Chinese and includes embedded English directives such as “Human owns verification,” but it does not state whether the skill should operate in Chinese, English, or adapt to the user's preferred language. This can amount to an implicit language constraint without user opt-in.

Content

No source excerpt is available for this finding.