Boost Skills Laravel Package

Agents

This is a public, open-source repository. The skill bodies in resources/boost/skills/, the guideline files in resources/boost/guidelines/, and README.md / UPGRADING.md ship in the Composer dist archive and are world-readable on GitHub. CHANGELOG.md, plans/, and .ai/ are export-ignored, but that...

claude-code

cursor

ai-guidelines

MUST USE when creating, updating, or managing AI skills and guidelines. Activates when: creating a new skill, editing existing skills, updating coding guidelines, modifying AI instructions, working with .ai/ directory, or when user mentions: skill, CLAUDE.md, guidelines, AI configuration.

claude-code

autoresearch

Autonomous performance optimization loop. Iteratively reduces query count and execution time by modifying code, benchmarking against measured noise, and keeping/reverting changes — then distills the kept changes down to the ones worth their complexity. Activates when: optimizing performance, reducing overhead, improving execution time, benchmarking, or when user mentions: autoresearch, optimize, performance, benchmark.

claude-code

backend-quality

Runs backend code quality checks in two tiers: Pint + related tests (every change), PHPStan + full suite (completion only). Activate after making changes to PHP files, or when user mentions: phpstan, pint, code quality, static analysis, code style, run checks.

claude-code

bug-fixing

Test-driven bug fixing workflow: reproduce with a failing test, find the root cause, then fix it. Activates when: fixing bugs, debugging issues, resolving defects, investigating errors, a failing or flaky test, or when user mentions: bug, fix, broken, not working, error, issue, defect, regression, flaky, root cause.

claude-code

clarify

Turns a fuzzy ask into sharp, fact-checked intent before any work starts — reduces ambiguity, closes gaps, sharpens overloaded terms, and surfaces hidden assumptions, fact-checking against the codebase instead of asking what the code can answer. The shared questioning core that interview and promptimize build on; also usable standalone. Activates when: clarifying a vague request, disambiguating intent, reducing ambiguity, sharpening fuzzy terms, running a gap analysis, surfacing assumptions, or when user mentions: clarify, disambiguate, sharpen, vague, ambiguous, gaps, gap analysis, unclear ask.

claude-code

clean-specs

Command-only — invoke with `/clean-specs`; never activates on its own. Removes spec files under the project's spec directories whose work is fully implemented and on the base branch, so the folders hold only live work. Fully implemented = every task checkbox checked AND the work proven on the base branch — by the implementation being present there, or by a merged PR. A spec still in an open/draft PR is NOT done and is kept.

claude-code

code-review

Structured review of recent code changes across functionality, UX/UI, design, code quality, security, testing — produces a prioritized findings list. Activates when: the user asks to review implementation, review changes, review code, audit code, check for improvements, or when user mentions: review, audit, improvements, code quality, code review.

claude-code

codex-review

Requests an independent code review from OpenAI Codex CLI, critically evaluates its findings, applies warranted fixes, and re-reviews until clean. Activates when: the user says /codex-review, asks for a Codex review, or wants an external AI review of changes.

claude-code

conventional-commits

Choose the Conventional Commits type, scope and breaking marker for a commit or PR title, and set up the CI check that rejects a non-conforming PR title. Activates when: writing a commit message or PR title in a Conventional Commits project, a title check failed, or when user mentions: conventional commits, commit type, feat or fix, breaking change, PR title check, semantic PR.

claude-code

deploying-laravel-cloud

Deploys and manages Laravel applications on Laravel Cloud using the `cloud` CLI. Use when the user wants to deploy an app, ship to cloud, create/manage environments, databases, caches, domains, instances, background processes, check billing/usage/spend, or any Laravel Cloud infrastructure. Triggers on deploy, ship, cloud management, environment setup, database provisioning, billing/usage queries, and similar cloud operations.

claude-code

eloquent-models

Creates and maintains Eloquent models with strict conventions: column/relation constants, comprehensive class docblock with @property and @property-read sections, foreign keys referenced via constants, and casts/fillable/hidden referencing constants. Use when creating a new model, adding columns or properties to an existing model, adding or modifying relations, or updating model docblocks. Triggers on model, eloquent, relation, belongs to, has many, has one, belongs to many, model constants, model properties, eloquent casts.

claude-code

evaluate

Self-directed eval loop: review own implementation, fix issues found, re-evaluate until clean, then run code review. Fix yourself; only ask the user for decisions. Activates when: evaluating implementation, self-reviewing code, checking for issues, or when user mentions: evaluate, check implementation, self-review, verify implementation.

claude-code

eye-verification

Command-only — invoke with `/eye-verification`; never activates on its own. Drives a frontend change in a real browser as the whole job, not one step of a quality gate: resolves the testables (PR body, tracker issue, spec edge cases, or gathered from the diff), executes each one against a running app, then stores the proof screenshots durably and attaches them to the PR/issue that exists. Reports Pass / Fail / NOT VERIFIED per testable with evidence.

claude-code

final-verification-review

Closeout verdict before shipping: runs the full evaluate loop (incl. Codex review), then dry-runs the closeout preflight check-only — the pull-requests preflight + gates in a PR flow, or the direct commit/release preconditions in a no-PR flow — and reports READY or NOT READY with what's missing. Creates nothing. Activates when: work is done and a commit/PR/release is next, final verification, final check, closeout check, or when user mentions: final-verification-review, final review, ship check, ready to ship, ready for PR, ready to commit.

claude-code

frontend-quality

Frontend quality checks: TypeScript type check, linting, and the JS test suite (Vitest/Jest). Activate after editing JS/TS files. Triggers: eslint, typescript, tsc, type check, lint, vitest, jest, test:js, frontend tests, frontend checks.

claude-code

github-issue-updates

Appends post-implementation content to a GitHub issue after a feature is built — a user-facing description and QA testables — and moves the issue on its project board. Activates when: updating a GitHub issue, documenting a feature for QA, writing an issue description, moving an issue's status, or when user mentions: issue update, github issue, testables, QA checklist, feature documentation.

claude-code

humanizer

Removes signs of AI-generated writing — staged contrasts, one-line closers, inflated significance, sales language, AI vocabulary, edit-summary filler, chatbot residue, document formatting in a reply — so prose a person reads is plain and clear, without adding or dropping facts. Returns only the final text when it gates a post, comment, commit or PR. Activates when: editing or reviewing prose a person reads, humanizing text, removing AI tells, polishing a README or release notes, or when user mentions: humanize, AI writing, sounds like AI, de-slop, writing review. NOT for machine-to-machine text — a message to another agent, a subagent prompt or its report, or any internal handoff needs no humanizing pass.

claude-code

implement-spec

Implements a specification file phase-by-phase with progress tracking. Activates when: implementing a spec, building from a spec, starting a spec phase, or when user mentions: implement spec, spec file, implement phase, build spec, start phase.

claude-code

interview

Adversarial grilling flow that gathers requirements for a complex feature by challenging assumptions, sharpening fuzzy terms, and cross-referencing the codebase before handing off to write-spec. Pairs with write-spec: interview gathers the requirements, then write-spec turns them into the spec file. Activates when: gathering requirements, planning a feature, clarifying ambiguous work, or when user mentions: interview, grill, requirements gathering, feature planning, gather requirements, ask me about.

claude-code

jira-create

Creates a new Jira issue with a well-formed, user-facing description. Triggers: create/open/file/log a Jira issue, make a ticket, file a bug, log a task, bug report. May activate for AI-identified follow-up work, but only after the user confirms — never silently. NOT for updating existing issues or Blocked-by-Question (use jira-updates), or rework research (use jira-rework).

claude-code

jira-rework

Researches a Jira issue sent back for rework: reads the issue, comments, and codebase to understand QA/UX feedback, then proposes fix options. Triggers: rework, rework issue, QA feedback, QA issue, investigate issue, research issue, UX feedback from Jira.

claude-code

jira-updates

Updates a Jira issue's description after its PR is created, and posts Blocked-by-Question comments. Invoke after creating a PR for a Jira issue. Triggers: PR just created for Jira issue, jira update, testables, QA checklist, translation strings, blocker comment, blocked by question. NOT for new issues (use jira-create) or rework research (use jira-rework).

claude-code

migration-squash

Conventions and verification checklist for creating or reviewing a Laravel migration squash (schema:dump --prune into a single schema baseline) — catches an incomplete, contaminated, or data-losing dump before it breaks fresh installs. Invoke-only: run only when explicitly invoked (e.g. /migration-squash) or when the user directly asks to create or review a migration squash. Do NOT auto-activate on incidental mentions of migrations, schema:dump, baselines, or pruning.

claude-code

php-generics

PHP docblock generics and shape conventions for static analysis: name a repeated array shape with @phpstan-type, import it across files, bind a generic base class with @extends / @implements / @use, and type a class-name parameter as class-string. Activates when: writing or tightening a docblock, a fixed-shape array return, a class-name parameter, a subclass of a generic base, a repeated array{...} shape, a Pest dataset closure, or when the user mentions: generics, @template, @extends, @use, @implements, @phpstan-type, @phpstan-import-type, class-string, array shape, type coverage.

claude-code

pre-release

Pre-push / pre-release checklist. Runs Rector, Pint, full test suite, PHPStan, and audits README, the docs site (docs/ pages, index sync, local build), and `.ai/` docs for staleness. Activate before: pushing to remote, tagging a release, writing release notes, or when user mentions: pre-release, pre-push, release checklist, ship, cut release, release notes.

claude-code

promptimize

Turns a rough, vague prompt into one optimized, model-agnostic prompt — closes gaps and ambiguities, fact-checks referenced material against the codebase, then rewrites using durable prompt-engineering principles and returns only the final prompt. Activates when: optimizing a prompt, improving a prompt, making a prompt clearer, tightening an instruction, prompt-engineering help, or when user mentions: promptimize, optimize prompt, improve prompt, better prompt, rewrite prompt, prompt engineering.

claude-code

pr-review-feedback

Applies PR review feedback with critical evaluation, then replies to and resolves the threads. Covers automated reviewers (Copilot, CodeRabbit, etc.). Activates when applying review comments, addressing PR feedback, responding to code review, or resolving review threads — including terse and Dutch phrasings: fix comments, fix the comments, fix PR comments, fix review comments, fix comments issue 1234, resolve/mark comments resolved, verwerk de comments, comments fixen. A bare number is resolved to the PR or its linked issue (Phase 0).

claude-code

prune-tests

Command-only — invoke with `/prune-tests`; never activates on its own. Audits a PHPUnit or Pest suite for low-value tests (framework restatements, echoes, exact duplicates, redundant data-provider rows, tests of deployed migrations and one-offs), has an independent refuter check every proposed deletion, removes the approved tests with a companion script, and proves the result against a baseline before it opens a cleanup PR.

claude-code

pull-requests

Creates and manages your own GitHub PRs via the gh CLI — analyze commits, write the PR description, create, verify, request review, route by risk. Activates when: creating or opening a PR, submitting work for review, writing or updating a PR description, or when user mentions: create PR, open PR, submit for review, write PR description, update PR, pull request. NOT for applying reviewer comments — use pr-review-feedback.

claude-code

readme

Author and maintain a concise, high-quality README for a Composer package. Covers stub, comprehensive, and docs-site shapes (thin README + a docs/ site, for example VitePress on GitHub Pages), length budgets, curated coverage, voice, the agent-facing surface (llms.txt, llms-full.txt, per-page markdown, robots.txt), staleness audits, and docs index/link sync. Activates when: writing or auditing a README or a docs page, or when the user says the docs are too long, too verbose, over-explained, read like a feature list, or asks to shorten or trim them.

claude-code

release-notes

Draft GitHub release bodies for Composer packages. Covers structure, length budget, voice, breaking-change callouts, and what to omit. Activates when release notes read too long, too verbose, or over-explained.

claude-code

release-please

Set up, run and repair release-please releases in a PHP repository on GitHub: bootstrap the config and manifest, choose the token, check the release PR, hand off the merge, and fix a missing, unwanted or wrongly versioned release. Activates when: setting up release-please, a release PR is open, cutting a release in a release-please repo, a change is missing from a release, an unwanted release PR opened, or when user mentions: release-please, release PR, autorelease, Release-As.

claude-code

resolve-conflicts

Resolves git merge conflicts without dropping functionality from either side. Merge is the main flow; rebase and stacked branches (gh stack) covered as auxiliaries. Activates when: merging branches, resolving conflicts, updating a branch, integrating work, rebasing a stack, or when user mentions: merge conflict, resolve conflicts, rebase conflict, integrate branch, update branch, stacked PRs, gh stack, update the stack.

claude-code

simplify-shape

Judge whether a change carries its values in the right type — a fixed set of strings that wants an enum, inline request validation that wants a form request, an array shape that wants a DTO, a repeated query chain that wants a query-builder method. Activates when: asking whether something should be an enum, a DTO or a form request, reshaping a value, or when the user mentions: primitive obsession, extract a form request, value object, array shape, magic strings, wrong type.

claude-code

test-value

Judges the tests a change added or changed, in both directions: the ones that prove nothing, and the behaviour with no test at all. Applies delete / rewrite / keep verdicts and covers the real gaps. Activates when: reviewing a change's tests, trimming a suite, asking whether a test is worth keeping or what is untested, or when user mentions: test value, useless tests, do these tests prove anything, trim tests, missing assertions, test coverage gaps.

claude-code

test-writing

Test-writing conventions for clear, targeted tests. Activates when: writing a new test, creating a test file, adding test coverage, naming tests, or when user mentions: test, testing, test case, coverage, assertion, unit test, feature test.

claude-code

upgrading

Canonical structure for UPGRADING.md in a Composer package. When to maintain one, what to put in it.

claude-code

ux-review

UX/UI expert review for new features and novel UI — evaluates the current state, explores options, then recommends and documents a chosen approach. Activates when: designing a new user-facing feature, adding a novel interaction pattern, redesigning an interface or user flow, or when user mentions: UX, UI, user experience, interface design, user flow, interaction design, usability, ux-review.

claude-code

write-spec

Writes implementation-ready specification files with progress-trackable phases. Activates when: writing a spec, creating a spec file, documenting a feature plan, or when user mentions: write spec, create spec, spec format, spec template.

claude-code