Agent Readiness Kit — Security skill for Claude Code

agent-readiness-kit

A deterministic, local-first CLI that checks whether a repository is ready for AI coding agents — Cursor, Claude Code, Codex, GitHub Copilot, and other coding agents.

It inspects repo structure, configuration, documentation, scripts, tests, and agent instruction files, then produces a terminal summary plus optional JSON, Markdown, or HTML reports. Use `ark init`, `ark generate`, and `ark fix` to scaffold practical starter files.

It runs entirely on your machine: no telemetry, no network calls, no LLM calls. It's meant to make a repo safer and easier for agents to work in, and to complement human review — not replace it.

Install

`ark` is not published to npm yet. **The npm package named `agent-readiness-kit` is an unrelated project**; installing it does not give you this tool and runs someone else's code. Install from GitHub instead (the CLI is built during install):

npm install -g github:alipajand/agent-readiness-kit
# or, as a dev dependency pinned to a reviewed commit
pnpm add -D github:alipajand/agent-readiness-kit#

Or clone and build locally:

pnpm install
pnpm build

Global use after build:

pnpm link --global
ark audit

Local development:

pnpm install
pnpm dev audit

Commands

Audit

Command	Description
ark audit	Run audit, print terminal summary, record score history
ark audit --json	Machine-readable JSON on stdout
ark audit --junit	JUnit XML on stdout (CI integration)
ark audit --sarif	SARIF 2.1.0 JSON on stdout (GitHub code scanning)
ark audit --output report.md