Back to skill

Security audit

Agent Pool

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent agent registry, but it includes persistent shared knowledge storage and outbound LinkedIn/email actions without clear approval boundaries.

Review this skill before installing if you plan to let it drive live marketing workflows. Require explicit approval for LinkedIn posts, comments, email sends, and customer research, and define what data may be written into shared knowledge-base and MEMORY.md files.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (17)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill metadata declares broad triggers such as querying the agent list, orchestrating new flows, and onboarding new agents without clear guardrails, which can overlap with ordinary user requests. In an agent-pool registry that influences orchestration, ambiguous activation increases the chance of unintended invocation or misuse of registry behavior during unrelated conversations.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
78% confidence
Finding

该技能文件整体以中文固定描述 Agent 名称、能力和触发方式,没有说明是否支持用户选择其他语言或是否仅限特定中文业务环境。根据语言/locale 政策,若技能默认强制单一语言而无用户选择或明确地域性理由,可能构成自然语言策略问题。

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Many per-agent trigger fields use vague conditions like 'CMO调度', '系统流程调用', or '完成后触发' without defining authorization boundaries, input predicates, or conflict resolution. Because this file is a global registry for multiple agents, ambiguous triggers can cause the wrong agent to be selected or multiple agents to activate inappropriately, expanding access to tools, workspaces, and memory paths.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The Researcher agent is explicitly described as writing collected data into a local knowledge base and shared directories, with persistence and cross-system readability, but the registry provides no user-facing warning, approval boundary, or restriction on what may be stored. In an agent-pool registry, this is risky because other workflows may invoke the agent assuming it is read-only, leading to unintended data modification, retention of sensitive information, or propagation of tainted content across shared memory paths.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The channel agents are defined to publish LinkedIn posts, perform comment interactions, and send email sequences, which are external-facing actions that can directly affect third parties and the organization's reputation. Because the registry lacks explicit warnings, approval requirements, or action boundaries, an orchestrator or user may invoke these agents without realizing they can autonomously perform outreach, creating risk of spam, unauthorized communications, or reputational and compliance harm.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 82)May include surrounding context.

md
# 系统接入模板 · 新增业务系统标准化规范

> 新增任何业务系统时,按此模板创建
> 参考:skills/agent-pool/SKILL.md(Agent池)+ skills/tool-pool/SKILL.md(工具池)
> 最后更新:2026-04-11

---

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/system-template.md (reported line 4)May include surrounding context.

md
# 系统接入模板 · 新增业务系统标准化规范

> 新增任何业务系统时,按此模板创建
> 参考:skills/agent-pool/SKILL.md(Agent池)+ skills/tool-pool/SKILL.md(工具池)
> 最后更新:2026-04-11

---

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/system-template.md (reported line 71)May include surrounding context.

md
# 系统接入模板 · 新增业务系统标准化规范

> 新增任何业务系统时,按此模板创建
> 参考:skills/agent-pool/SKILL.md(Agent池)+ skills/tool-pool/SKILL.md(工具池)
> 最后更新:2026-04-11

---

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/system-template.md (reported line 86)May include surrounding context.

md
# 系统接入模板 · 新增业务系统标准化规范

> 新增任何业务系统时,按此模板创建
> 参考:skills/agent-pool/SKILL.md(Agent池)+ skills/tool-pool/SKILL.md(工具池)
> 最后更新:2026-04-11

---

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/system-template.md (reported line 140)May include surrounding context.

md
# 系统接入模板 · 新增业务系统标准化规范

> 新增任何业务系统时,按此模板创建
> 参考:skills/agent-pool/SKILL.md(Agent池)+ skills/tool-pool/SKILL.md(工具池)
> 最后更新:2026-04-11

---

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/system-template.md (reported line 157)May include surrounding context.

md
# 系统接入模板 · 新增业务系统标准化规范

> 新增任何业务系统时,按此模板创建
> 参考:skills/agent-pool/SKILL.md(Agent池)+ skills/tool-pool/SKILL.md(工具池)
> 最后更新:2026-04-11

---

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/system-template.md (reported line 4)May include surrounding context.

md
# 系统接入模板 · 新增业务系统标准化规范

> 新增任何业务系统时,按此模板创建
> 参考:skills/agent-pool/SKILL.md(Agent池)+ skills/tool-pool/SKILL.md(工具池)
> 最后更新:2026-04-11

---

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/system-template.md (reported line 22)May include surrounding context.

md
# 系统接入模板 · 新增业务系统标准化规范

> 新增任何业务系统时,按此模板创建
> 参考:skills/agent-pool/SKILL.md(Agent池)+ skills/tool-pool/SKILL.md(工具池)
> 最后更新:2026-04-11

---

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/system-template.md (reported line 155)May include surrounding context.

md
# 系统接入模板 · 新增业务系统标准化规范

> 新增任何业务系统时,按此模板创建
> 参考:skills/agent-pool/SKILL.md(Agent池)+ skills/tool-pool/SKILL.md(工具池)
> 最后更新:2026-04-11

---

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The template’s example describes automatic generation of a customer follow-up report immediately after sales data entry, but it does not require any user notice, consent check, or data-handling disclosure. In this skill context, the document is a standard onboarding template, so omissions here are likely to propagate across multiple downstream systems and normalize silent processing of potentially sensitive customer information.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The example workflow explicitly invokes external research tools to collect customer background information without any user-facing warning, approval gate, or restriction on what customer identifiers may be sent to third parties. Because this is a reusable system template, it encourages future implementers to adopt external data enrichment patterns without privacy review, increasing the chance of unauthorized disclosure of customer-related data.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
72% confidence
Finding

SQP-3 applies to natural-language policy violations across all file types, including language or locale constraints. This template appears to require Chinese for all instructions and examples without indicating user choice, opt-in, or a documented region-specific justification.

Content

No source excerpt is available for this finding.