Security audit

Agent Message Bridge

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a real local agent messaging tool, but its file handling is under-scoped and can expose or silently delete messages.

Review before installing. Use only in a trusted local environment, do not send secrets or regulated data through it, restrict filesystem permissions on the message directory, and patch or configure the storage path and agent-name validation before relying on it for multi-agent workflows.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (4)

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The README advertises message history, polling, and file-based queues but does not disclose that messages are likely stored as plaintext on disk and may be continuously scanned by polling agents. In an inter-agent communication tool, this can lead users to place sensitive operational data into an insecure local mailbox model, increasing the risk of unintended disclosure, tampering, or forensic leakage.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The skill promotes direct inter-agent messaging, automatic polling, and complete message records, but does not warn users that message contents may be stored, replayed, or exposed to other agents or local users. In a communication tool, omission of privacy and retention disclosures is risky because users may send sensitive operational data under the assumption that the channel is ephemeral or private.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The professional and enterprise descriptions promote HTTP API real-time communication and web-based features without a clear warning that data may be transmitted over a network and exposed to interception, logging, or third-party infrastructure. Because the skill's purpose is cross-agent communication, users may transmit sensitive task data, making undisclosed network transport materially risky.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The bridge deletes inbox files by default immediately after reading them, with no confirmation, audit trail, or separate acknowledgement step. In an inter-agent messaging bridge, this can cause silent loss of evidence, missed instructions, and easy denial-of-service by triggering reads that consume messages before they are processed elsewhere.

Static analysis

No suspicious patterns detected.