HoneyBadger FAQs

Name: HoneyBadger
Author: famclaw

Question 1

What are 'paranoia levels' in HoneyBadger and how do they work?

Accepted Answer

Paranoia levels (off, minimal, family, strict, paranoid) allow users to tailor the scanner's enforcement based on their security needs. Each level activates a specific set of scanners and sets different thresholds for blocking findings based on severity, enabling flexible and targeted security policies.

Question 2

What is HoneyBadger and how does it secure AI agents?

Accepted Answer

HoneyBadger is a comprehensive security scanner for AI agent skills, tools, and MCP servers. It performs static analysis on source code and metadata to detect vulnerabilities like secrets, CVEs, supply chain attacks, and prompt injection risks *before* installation or runtime.

Question 3

Can HoneyBadger operate without an internet connection?

Accepted Answer

Yes, HoneyBadger is designed for full offline capability. It's a self-contained Go binary that can perform scans without requiring external network access or cloud dependencies, making it suitable for secure and isolated environments.

Question 4

What types of threats can HoneyBadger identify?

Accepted Answer

It identifies a wide range of threats, including exposed secrets, known CVEs in dependencies (across 8 lockfile formats), supply chain attacks (e.g., typosquatting, remote script execution), prompt injection vulnerabilities, permission mismatches, and issues in SKILL.md metadata.

Question 5

How does HoneyBadger integrate into development and deployment workflows?

Accepted Answer

HoneyBadger runs as a single, cross-platform Go binary, supporting CLI usage for CI/CD pipelines, a native MCP server mode via JSON-RPC for agent runtime integration, and direct integrations with AI platforms like Claude Code, OpenAI Codex CLI, FamClaw, OpenClaw, and NanoBot for pre-install gating.

HoneyBadger

HoneyBadger

主要功能

使用案例

主要功能

使用案例