Laravel Laravel Package

The instructions for coding agents in this repository are in CLAUDE.md. Read it in full before you change anything. It is the only copy: this file just points to it, so the two cannot drift apart.

Add or change an Agentic Actions action, wire it into a laravel/ai agent, add the copilot panel with live rows, expose actions over MCP with scoped tokens, connect a remote MCP client such as a Claude or ChatGPT connector with OAuth, let the copilot run a Destructive or External action after the person confirms, let it ask the person in a form for what a call left out, show a Read action's rows as a table, or let the model ask questions about one model's rows with a dataset. Use when a task touches AgenticActions\Action classes,

Invoke whenever writing, changing, reviewing, or sweeping tests in this package (Pest under tests/, node tests under js/tests/). Authoring gate for new tests plus the audit workflow for low-value, implementation-coupled, or duplicative tests and the test-only seams they keep alive.

Campaign mode prunes one area's whole test surface: one module (Streaming, OAuth), or the whole suite in lanes. The value bar, retention bar, candidate evidence and validation in SKILL.md apply to every lane. This file adds the order of work. Each step ends on its completion criterion; do not...

SKILL.md and CAMPAIGN.md are adapted from the test-audit skill of OpenClaw (https://github.com/openclaw/openclaw, .agents/skills/test-audit/ at commit 80930af), used under the MIT License: