AI-ready CLI

AI-ready is a local CLI that prepares an existing project for reliable AI Coding with Claude, OpenCode, Kimi, Codex, or similar tools.

It is not another coding agent. It builds the project readiness layer, project understanding layer, and workflow discipline layer around the coding agents your team already uses.

Install

npm install -g @fe178/ai-ready

Run it inside a project:

cd /path/to/your-project
air

The package exposes two commands:

air
ai-ready

What AI-ready Does

1. Understand This Project

air "理解业务"
air "建立业务地图"

AI-ready checks whether the project has enough usable material for a trustworthy business map:

• code entry points
• business documents
• OpenSpec changes or requirements
• API or architecture docs
• human-reviewed rules

If sources are missing, it shows the next material to add instead of inventing a fake map.

2. Prepare an AI Task

air "设置页保存后展示成功提示"

AI-ready creates .ai-ready/task-pack.md and related metadata so a coding agent can receive task-relevant context by file attachment instead of pasted prompt text.

3. Check an AI Result

For OpenSpec-style teams, AI-ready can produce verification and review evidence for a change:

air openspec select --project . --change <change-id> --task <task-id> --json
air openspec verify --project . --change <change-id> --json
air openspec accept --project . --change <change-id> --json
air openspec report --project . --change <change-id> --json

Local Artifacts

AI-ready writes local project artifacts under .ai-ready/ in the target project, for example:

.ai-ready/task-pack.md
.ai-ready/task-pack.json
.ai-ready/wiki/business-map.md
.ai-ready/wiki/business-map-readiness.json
.ai-ready/wiki/candidates.jsonl
.ai-ready/wiki/accepted-rules.yaml

These files can contain derived project information. Review them before committing.

Runtime Adapters

AI-ready can call local CLIs that are already installed and logged in on your machine:

air runtime doctor
air runtime list

The tool does not require model API keys in its own config for the default local CLI workflow.

Security Notes

• Do not add .env, private keys, cookies, tokens, or credential files as sources.
• Do not publish .ai-ready/wiki/runs, raw snapshots, screenshots, or logs unless reviewed.
• Model output is treated as candidate knowledge. It becomes trusted only after human review.
• Use private package registries for internal pilots if the package or docs reference internal workflows.

See SECURITY_NOTES.md for the detailed security model.

Requirements

• Node.js 20+
• macOS/Linux shell environment
• Optional local coding-agent CLIs: Claude Code, OpenCode, Kimi, Codex

Status

This is a 0.x pilot package. Schema and command details can still change before 1.0. The pilot npm package is published under the maintainer scope @fe178/ai-ready; the command name remains air, and the package can migrate to a formal org scope later.