Software Development Skills

Run a bounded architectural refactor loop. Use this when the user asks to improve architecture, refactor toward a stated design, untangle module boundaries, fix dependency direction, reshape code organization, or iterate until the architecture is satisfactory while live-testing, reviewing, checkpointing, and recording progress.

Audit a codebase, path, glob, or branch diff for classic code smells using the 56-smell Luzkan catalog, with detection heuristics tuned to PHP/Laravel and TS/React — sweep nine occurrence lenses, adversarially verify every candidate against the smell's card definition, emit a ranked markdown findings report. Use when the user says "audit for code smells", "code smell audit", "run code-smells-audit", "smell check", "smell-check this diff / this controller / this component", or asks which catalog smells a module, Laravel class, or React component exhibits.

Shared vocabulary for designing deep modules in a PHP/Laravel codebase. Use when the user wants to design or improve a class or module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill (to-spec-plus, implement-plus, architecture-satisfaction) needs the deep-module vocabulary.

Mine your recent AI-coding session logs (Claude Code + OpenAI Codex) for reusable patterns — corrections you gave, commands that errored or were retried, setup steps rediscovered across sessions, and content-worthy moments — then propose where each belongs (CLAUDE.md/AGENTS.md line, slash command/skill, hook, tool fix, config change, or content idea). Use when the user says "read my recent sessions", "distill my sessions", "what patterns are in my logs", "mine my history", or asks to turn their session history into rules/skills/content.

Build and sharpen a project's domain model — challenge terms against the glossary, resolve fuzzy language via AskUserQuestion, stress-test concepts with concrete scenarios, and record CONTEXT.md entries and ADRs the moment decisions land. Use when pinning down domain terminology or a ubiquitous language, recording an architectural decision, or when another skill (grill-me-plus, wayfinder-plus, to-spec-plus) needs the domain model maintained.

Interview the user relentlessly about a plan, decision, or idea until reaching shared understanding, resolving each branch of the decision tree. Use Claude Code's AskUserQuestion or Codex's requestuserinput when available so decisions are click-to-answer. Use when the user says "grill me", wants to stress-test their thinking, compare design choices, or resolve ambiguous implementation decisions.

Implement one ticket or spec end to end in a Laravel codebase — claim it, TDD at pre-agreed seams with Pest/PHPUnit, run static analysis and Pint, review and verify the result, commit, and close the ticket. Use when the user says "implement this ticket/spec", passes a ready-for-agent issue, or wants to work the next frontier ticket.

Use when working with Laravel projects served by Laravel Herd alongside git worktrees on macOS. Triggers on "laravel herd worktree", "new worktree for laravel", "herd parked sites worktree", "multiple branches herd", "composer install per worktree", ".test url for branch", "isolate php version per worktree", or any task that creates, lists, or removes a git worktree inside a Herd-served Laravel project. Covers site routing, per-worktree .env/database/cache isolation, and fast dependency install strategy.

Run a documentation sweep for a codebase. Use this when the user asks for a nightly docs sweep, overnight docs sweep, documentation drift audit, README/setup/API docs refresh, docs sync after implementation changes, or a reviewable pull request that makes docs match the current code.

Audit and clean Git repository state. Use this when the user asks for repository cleanup, stale branch cleanup, old pull request review, unmerged commit or stash recovery, git worktree cleanup, remote pruning, abandoned work recovery, or a repo-state hygiene pass across branches, PRs, commits, stashes, and worktrees.

Investigate a question against high-trust primary sources and capture the findings as a cited Markdown file, run as a background subagent so the main session keeps…

Turn a decision the user can't fully answer into a Markdown questionnaire for someone else to fill in, async or over a meeting. Use Claude Code's AskUserQuestion or Codex's requestuserinput for the setup questions. Use when the user says "to questionnaire", needs answers from a stakeholder, vendor, or compliance contact, or a grilling branch is blocked on knowledge someone else holds.

Turn the current conversation into a spec (PRD) and publish it to the project's issue tracker — no interview, just synthesis of what was already discussed, with Laravel test seams confirmed via AskUserQuestion before writing. Use when the user says "to spec", "write this up as a spec/PRD", or a grilling session is done and needs a durable artifact.

Break a plan, spec, or the current conversation into tracer-bullet vertical-slice tickets with explicit blocking edges, approve the breakdown via AskUserQuestion, and publish to the project's issue tracker (GitHub, Linear, or local markdown). Use when the user says "to tickets", "break this into tickets/issues", or a spec needs slicing for implementation.

Plan work too big for one agent session as a shared map of decision tickets on the project's issue tracker (GitHub, Linear, or local markdown), then resolve them one per session until the route to the destination is clear. Use when the user says "wayfinder", "chart a map", "work the map", or brings a loose idea too large or foggy to spec in one sitting.

Run a weekly Composer dependency audit for PHP projects. Use this when the user asks for a Composer dependency audit, PHP dependency review, composer outdated package summary, composer audit, Laravel package upgrade plan, or a safe PHP dependency upgrade summary.

Run a weekly npm dependency audit for JavaScript or TypeScript projects. Use this when the user asks for an npm dependency audit, npm outdated package review, package.json upgrade plan, weekly dependency audit, npm security check, or a safe npm upgrade summary.