Laravel Boost v2.6.0 consolidates its previously fragmented testing skills (pest-testing, enforce-testing, phpunit-guidelines) into a single testing-best-practices skill that dynamically adapts to Pest, PHPUnit, browser testing, and Test Impact Analysis, covering nine guidance areas from assertions to security testing. The release also replaces lexical query parsing in the DatabaseQuery MCP tool with database-enforced read-only transactions (using SET TRANSACTION READ ONLY on MySQL/MariaDB/PostgreSQL and PRAGMA query_only on SQLite) to reliably block write operations, plus fixes for skill installation, MySQL ANSI_QUOTES compatibility, MCP JSON config formatting, and Antigravity agent detection.
Table of contents
# Testing Best Practices Skill# Database-Enforced Read-Only Transactions# Skill Management and Tooling ImprovementsQuestions this post answers
What testing skills does Laravel Boost v2.6.0 replace with a single unified skill?
Laravel Boost v2.6.0 replaces the previously separate pest-testing, enforce-testing, and phpunit-guidelines skills with one consolidated testing-best-practices skill. The overlap between the old skills could cause AI agents to produce inconsistent test suites, test internal framework details, or add redundant mocks. The new skill dynamically adapts based on installed testing packages, covering Pest, PHPUnit, browser testing, and Test Impact Analysis. daily.dev helps teams tuning AI-agent testing workflows keep up with Laravel Boost changes like this.
How does the DatabaseQuery MCP tool in Laravel Boost 2.6.0 prevent write queries from executing?
It runs every query inside a database-enforced read-only transaction rather than relying on lexical keyword filtering, since keyword checks can miss complex shapes like data-modifying CTEs. On MySQL and MariaDB it issues SET TRANSACTION READ ONLY before starting the transaction, on PostgreSQL it sets that after starting the transaction, and on SQLite it sets PRAGMA query_only = ON. The transaction is always rolled back afterward. Developers wiring AI agents into their databases can track MCP safety changes like this via daily.dev.