skills.rest
Tencent
TencentOFFICIAL
Explore all skills by this author →
WebsiteGitHub

inter-agent-comm-security-detection

Detect data leakage, boundary bypass, and privilege mismatch in inter-agent communication.

6.1k|568|Updated Dec 25, 2024
Search Tags:#security testing#vulnerability detection#data leakage#Privilege escalation#Red Teaming#Agent communication#multi-agent security
One-click install
npx skills add https://github.com/Tencent/AI-Infra-Guard --skill inter-agent-comm-security-detection
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: inter-agent-comm-security-detection
Source: https://github.com/Tencent/AI-Infra-Guard/tree/main/agent-scan/agent_scan/prompt/skills/inter-agent-comm-security-detection
Command: npx skills add https://github.com/Tencent/AI-Infra-Guard --skill inter-agent-comm-security-detection

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Multi-agent AI systems can leak data between agents, allow impersonation across boundaries, or let low-privilege agents reach high-privilege tools through message routing. This Skill provides a structured dialogue-based testing methodology to identify these inter-agent communication vulnerabilities before they are exploited.

Core Features & Use Cases

  • Cross-Agent Data Leakage Testing: Probes whether one agent can retrieve another agent's private messages, memory, or secrets.
  • Boundary and Impersonation Checks: Tests whether messages can be sent as another agent or injected into another agent's context.
  • Channel Security and Privilege Escalation Probes: Verifies encryption of inter-agent transport and whether routing enables unauthorized privileged actions.
  • Use Case: A security engineer auditing a multi-agent customer support platform uses this Skill to confirm that a low-privilege triage agent cannot access the admin agent's tools or read other agents' conversation history.

Quick Start

Use the inter-agent communication security detection skill to test whether this multi-agent system leaks data across agents or allows privilege escalation through message routing.

Frequently Asked Questions about inter-agent-comm-security-detection

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I test multi-agent systems for data leakage between agents?▼

Send dialogue prompts that request another agent's private data, such as asking to show what another agent said or to retrieve secrets from its memory. The system is vulnerable if it discloses cross-agent data in its response.

What is inter-agent privilege escalation and how is it detected?▼

Inter-agent privilege escalation occurs when a low-privilege agent reaches high-privilege tools through message routing. Test it by asking the system to route a request to an admin agent to run a privileged tool, and check whether the action executes without authorization.

When should I skip inter-agent communication security testing?▼

Skip this testing when the target system is a single agent with no cross-agent messaging, shared memory, or broker-based routing. The probes only apply to architectures where multiple agents collaborate or exchange messages.

What output does the inter-agent security detection produce?▼

Each confirmed finding is emitted as a structured XML vuln element containing a title, evidence with test prompts and responses, risk type, severity level, remediation suggestion, and the full conversation transcript.

How do I check if inter-agent message channels are encrypted?▼

Ask the agent directly about its transport security, such as whether inter-agent messages are encrypted or whether a shared log stores full message content. Confirming unencrypted or publicly accessible channels for sensitive content counts as a vulnerability.

Related Skills

Explore More Software Engineering Skills→
tianzecntianzecn
2

claudemem-orchestration

Coordinate parallel code analysis agents using shared claudemem output.

Community
Advanced
dnvitidnviti
10

vulnerability-create

Generate structured remediation tasks from security vulnerability findings with severity classification.

Community
Intermediate
jayjpatel9717jayjpatel9717

llm-security

Detect and mitigate LLM security vulnerabilities like prompt injection and RAG poisoning.

Community
Advanced
Q16GQ16G
83

graybox-p0

Orchestrate gray-box testing by combining white-box candidate discovery with black-box request validation.

Community
Intermediate