Agentic Development

Andurel is built so humans and agents share one project pad: generated application code you own, an agent-ready CLI, and an installable skill.

Install the skill

1andurel skill show
2andurel skill install --harness claude,codex

skill install writes the embedded Andurel skill into the chosen harnesses (claude, codex, pi, opencode, crush). Prefer the skill over inventing paths, URLs, or generator names. Automation must pass --harness so installation never waits for interactive selection.

Project guidance

Generated applications include AGENTS.md. Agents should:

Run andurel commands --json before guessing subcommands
Prefer andurel inspect … --json for read-only discovery
Use --dry-run / structured output on mutating generators when available
Import routes from @/routes after andurel sync routes in Inertia apps

1andurel commands --json
2andurel inspect project --json
3andurel doctor --json
4andurel generate scaffold Product --dry-run --json
5andurel sync queries --json
6andurel inspect routes --json

Structured CLI output

Global flags make command results machine-readable:

Flag	Purpose
`--json`	JSON response envelope
`--md`	Markdown where supported
`--agent`	Structured agent-oriented output
`--jq`	Select a simple field path from command data
`--ids-only`	One resource identifier per line when supported
`--count`	One raw resource count when supported
`--quiet`	Suppress non-essential human progress

See Agent Output for mutation previews and Overview for the full command surface.