Security audit

A2A Chatting

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed helper for managing OpenClaw agent-to-agent chat sessions, with local session records and no hidden remote execution or privilege behavior found.

Install only if you want your agent to coordinate with other OpenClaw agents. Treat messages sent through this skill as shared with another agent context, avoid including secrets or unnecessary private data, and be aware that session metadata is stored locally until deleted.

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

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 (4)

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The skill's trigger conditions are very broad, covering nearly any situation involving another agent, coordination, or capability queries. In an agentic environment, this can cause unintended invocation and transmission of user or system-derived content to other agents without sufficient user awareness or narrowing, increasing the risk of unnecessary data sharing and cross-agent prompt injection exposure.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill instructs the agent to send messages to other agents and notes that session data is stored on disk, but it does not include a user-facing warning or consent mechanism. This creates a real privacy and security risk because users may unknowingly have their content forwarded to external agent contexts and retained in persistent storage, potentially exposing sensitive information.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: This code deletes the session file and rewrites the session index, which is a destructive operation affecting stored user data. Although a success message is printed afterward, there is no pre-action confirmation, cautionary notice, or inline warning comment/docstring disclosing that the command permanently removes session records.

Natural-Language Policy Violations

Low

Confidence: 85% confidence
Finding: The required message examples are written in Chinese and the document does not state that the language should follow user preference or allow locale choice. This can be read as prescribing a specific language for agent-to-agent communication without user opt-in.

Static analysis

No suspicious patterns detected.