AgentFence v0.2.0

AgentFence v0.2.0 is the scoped npm release candidate for @adulph3/agentfence. The executable remains agentfence. Node.js 22 and 24 are supported; Node.js 24 is recommended. The six Ubuntu, macOS, and Windows CI jobs and CodeQL passed on release commit 36d1a7f.

Scoped Quick Start after npm publication:
npx @adulph3/agentfence scan
npx @adulph3/agentfence scan --user-configs

Local-first configuration and instruction scanning remains non-executing, with no application-initiated network requests, telemetry, or backend during scan and doctor. The unrelated unscoped npm package agentfence is not this project.

Artifact: adulph3-agentfence-0.2.0.tgz
SHA-256: ac7c7bbcedcb07b1a290229ec417a353ce006f88d1efbd81f3184ecb30bd7429

AgentFence analyzes observable configuration and exposure. It does not guarantee that an AI agent, repository, or machine is secure.

AgentFence v0.1.0

AgentFence v0.1.0

AgentFence is a local-first security scanner for AI-agent configuration and instruction files.

• Supports Codex, Claude Code, Cursor, Kiro, and MCP configurations.
• Analyzes local configuration and instruction files without executing scanned content.
• Detects shell and supply-chain risks, secret-bearing environment variable names, Unicode/bidi hazards, and filesystem containment issues.
• Provides explainable security scoring, structured JSON output, and a doctor command.
• Runtime scanning uses zero network access, with no telemetry and no backend.

Official artifact:

• agentfence-0.1.0.tgz
• SHA-256: 4d5271aad1a7f56f66555752d6a3184643cac0213a712079af33d95709db278b

AgentFence analyzes observable configuration and exposure and does not guarantee that an AI agent, repository, or machine is secure.