agent-creator
Use when creating expert agents. Generates agent.md with frontmatter, hooks, required sections, and skill references.
Use when setting up PHPStan, Rector, or PHP-CS-Fixer on a non-Laravel PHP project, incl. CI wiring. Do NOT use for Laravel (Pint/Larastan), tests, or syntax.
$ npx -y skills add fusengine/agents --skill php-quality-tooling --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/php-quality-toolingContext preview
The summary Claude sees to decide when to auto-load this skill.
Use when setting up PHPStan, Rector, or PHP-CS-Fixer on a non-Laravel PHP project, incl. CI wiring. Do NOT use for Laravel (Pint/Larastan), tests, or syntax.
name: php-quality-tooling description: Use when setting up PHPStan, Rector, or PHP-CS-Fixer on a non-Laravel PHP project, incl. CI wiring. Do NOT use for Laravel (Pint/Larastan), tests, or syntax. versions: phpstan: "2.x" rector: "2.x" php-cs-fixer: "3.x" user-invocable: true references: references/phpstan-levels.md, references/rector-upgrades.md, references/php-cs-fixer.md, references/templates/phpstan-neon.md, references/templates/rector-php.md, references/templates/php-cs-fixer-dist.md, references/templates/composer-ci.md related-skills: php-testing, php-language-modern
<objective> Covers the three complementary quality tools for framework-agnostic PHP: PHPStan for static analysis (finds type bugs, never edits), Rector for automated upgrades and refactors (rewrites source), and PHP-CS-Fixer for coding-standard formatting (rewrites whitespace/style). They do not overlap — a full setup runs all three.
Includes a decision guide for greenfield vs legacy codebases (PHPStan level + baseline strategy, Rector's withPhpSets()/withPhpLevel(), PHP-CS-Fixer rule sets), plus ready-to-copy templates for phpstan.neon, rector.php, .php-cs-fixer.dist.php, and a composer-scripts CI pipeline.
Do NOT use this skill on Laravel projects — Laravel uses Pint (a thin wrapper over PHP-CS-Fixer) and Larastan instead, both owned by laravel-expert. Do NOT use it for test framework setup (php-testing) or language syntax questions (php-language-modern). </objective>
Three complementary tools. PHPStan **finds** type bugs, Rector **rewrites** code (upgrades + refactors), PHP-CS-Fixer **formats** to a coding standard. They do not overlap — run all three.
Before ANY implementation, spawn 3 agents in parallel, one `Agent` call each with a `name`:
1. **fuse-ai-pilot:explore-codebase** - Detect existing config (phpstan.neon, rector.php, .php-cs-fixer.dist.php), PHP version in composer.json 2. **fuse-ai-pilot:research-expert** - Verify latest PHPStan/Rector/PHP-CS-Fixer docs via Context7/Exa 3. **mcp__context7__query-docs** - Check current config API surface
After implementation, run **fuse-ai-pilot:sniper** for validation.
---
| Tool | Role | Config file | Verdict | |------|------|-------------|---------| | PHPStan | Static analysis (type safety, dead code) | `phpstan.neon` | Reports errors, never edits | | Rector | Automated upgrade + refactor | `rector.php` | Rewrites source | | PHP-CS-Fixer | Coding-standard formatter | `.php-cs-fixer.dist.php` | Rewrites whitespace/style |
Scope note: **Laravel Pint** is a thin opinionated wrapper around PHP-CS-Fixer. On a Laravel project use Pint (→ laravel-expert); everywhere else use PHP-CS-Fixer directly, which is what this skill covers.
---
1. **Pick one style tool** - PHP-CS-Fixer OR Pint, never both on one repo 2. **Rector runs `--dry-run` first** - Review the diff before applying; commit before a bare run 3. **Baseline, don't lower the level** - Adopt a high PHPStan level + baseline the debt, not `level 4` 4. **Style before analysis in CI** - Fixer `check` → PHPStan → tests; a formatting diff should fail fast 5. **Pin tool versions** - `--dev` in composer.json; a minor bump can add rules and break CI
---
New project from scratch? ├── Static analysis → PHPStan level 8-9+ from day one (our default; nothing to baseline yet) ├── Style → PHP-CS-Fixer @PER-CS (always-latest) or @PER-CS3.0 (pinned) └── Rector → withPhpSets() (reads composer.json) + deadCode/codeQuality prepared sets Legacy codebase? ├── PHPStan → start at a level that passes, generate baseline, raise 1 level per PR ├── Rector → withPhpLevel(n) / withTypeCoverageLevel(n), one level at a time └── Style → @PER-CS + fix once, then enforce check in CI
---
| Topic | Reference | Load when | |-------|-----------|-----------| | PHPStan levels & baseline | `references/phpstan-levels.md` | Choosing/raising analysis strictness | | Rector sets & upgrades | `references/rector-upgrades.md` | Automating PHP upgrades or refactors | | PHP-CS-Fixer rule sets | `references/php-cs-fixer.md` | Configuring code style |
| Template | Use Case | |----------|----------| | `references/templates/phpstan-neon.md` | Complete `phpstan.neon` (strict + baseline) | | `references/templates/rector-php.md` | Complete `rector.php` (upgrade + quality) | | `references/templates/php-cs-fixer-dist.md` | Complete `.php-cs-fixer.dist.php` (@PER-CS3.0) | | `references/templates/composer-ci.md` | composer scripts + GitHub Actions pipeline |
---
composer require --dev phpstan/phpstan rector/rector friendsofphp/php-cs-fixer vendor/bin/php-cs-fixer check # style gate (no writes) vendor/bin/phpstan analyse # type analysis vendor/bin/rector process --dry-run # preview refactors
→ Full config in `references/templates/`
---
A plugin ecosystem that turns Claude Code into a supervised, multi-agent development environment.
Repo: fusengine/agents
Use when creating expert agents. Generates agent.md with frontmatter, hooks, required sections, and skill references.
Use when starting ANY development task -- feature, bug fix, refactor, hotfix (triggers: implement, create, build, fix, add feature, refactor, develop).
Use when creating a feature/component or adding functionality. Fires BEFORE APEX Analyze to refine requirements via structured questioning.
Use before a root-cause, done/verified claim, irreversible action, or 2nd-time fix reaches the owner (APEX or plain conversation); also fires at every…
Use when validating code quality after modifications -- SOLID compliance, DRY duplication, linter errors, architecture violations. Do NOT use for functional…
Use when an expert agent self-reviews and self-corrects code after the Execute phase, before sniper validation (BMAD-METHOD elicitation techniques).