skills

Automated browser testing of UI flows via Playwright MCP + agent-browser. Use when user says "browser qa", "test the flow", "verify the UI", "test this page", "check if it works", "end-to-end check", or after completing code changes that affect UI and need manual verification replaced by automation. NOT for conversational bug-reporting sessions — that's /qa.

Use when a GitHub ticket/issue bundles multiple sub-tasks or fixes that will take hours, span context compactions, and ship as one reviewed PR — or when the user asks for subagent-driven execution, "the 4875 workflow", or proper progress tracking in the ticket.

Pre-code quality gate that runs /review-pr's anti-slop lens against a written plan BEFORE any code is written. Grounds the plan against the real codebase, runs 11 category checks (security, concurrency, round-trip, control-flow, error-handling, pattern-consistency, plus /review-pr's Q1-Q6), then grills the user one question at a time until the plan is hardened. Use when the user says "harden my plan", "check my plan", "grill my plan before I code", "lint this plan", or invokes `/harden-plan` explicitly. Also invoke proactively after `/grill-me` or `/grilling` completes with a written plan and before any implementation begins. Do NOT invoke after coding has started — redirect to `/review-pr` / `/fix-pr-review` in that case.

The disciplined root-cause loop to follow once you are already IN a debugging session — four phases (root-cause investigation, pattern analysis, hypothesis testing, implementation), no fix without an understood cause, evidence discipline at every component boundary, and a hard stop after repeated failed fixes. Use when about to propose or apply a fix for a bug or test failure, when a previous fix didn't work, when tempted to patch a symptom (add a null check, wrap in try-catch, bump a timeout, add a retry), or when you catch yourself guessing instead of tracing. Pairs with /diagnosing-bugs (repro + feedback-loop construction) — this skill governs fix discipline once the session is underway.

Execute git commit with conventional commit message analysis, intelligent staging, and message generation. Use when user asks to commit changes, create a git commit, or mentions "/commit". Supports: (1) Auto-detecting type and scope from changes, (2) Generating conventional commit messages from diff, (3) Interactive commit with optional type/scope/description overrides, (4) Intelligent file staging for logical grouping

Use when you need to resolve an in-progress git merge/rebase conflict.

Audit a stale GitHub issue/ticket against the current codebase — what's done, what's left, is it still needed? Then update or sunset it. Use when user says "audit this ticket", "is this ticket still needed", "update this old ticket", "what's done and what's left on this issue", or "should we sunset this issue".

MANDATORY post-task verification. Run after EVERY task — no exceptions, no skipping, regardless of task size. Executes type-check, parallel code review, and code simplification in sequence.

Triage and fix CodeRabbit / review-pr findings on a GitHub PR, then reply + resolve conversations. Classifies each finding (FIX/DISMISS/DEFER/DISAGREE/NEEDS-INPUT), runs /done, posts specific replies. Use on CodeRabbit review URLs, PR URLs, or local review files. Pairs with /review-pr.

Fix TypeScript errors in changed or specified files. Use when user says "ts errors", "typescript errors", "fix types", "type errors", "this file has ts error", "still not working" (in TS context), or after writing code that may have type issues. Also use proactively after editing TypeScript files.

Deep, anti-slop review of a GitHub PR. Grounds findings in the linked issue's intent, runs a Claude reviewer (+ conditional silent-failure hunter) in parallel with existing CodeRabbit comments fetched from the PR, then critic-passes the findings before printing. Use when user says "review this pr", pastes a GitHub PR URL, or asks "check this pull request". Also handles multiple PRs or "review all open PRs" via batch mode — one subagent per PR, consolidated report, decisions deferred to the end. NOT for local uncommitted changes — use /parallel-review for those.

Search the codebase before writing any new utility, type, schema, component, hook, constant, or helper — 3-layer search (name, behavior, reference), reuse hierarchy, and duplicate smells. Use BEFORE creating any new function, type, zod schema, component, hook, or constant; when tempted to add a file to utils/ or lib/; when about to copy-paste similar code; or when reviewing a diff for duplication. Triggers on "new utility", "create a helper", "add a type", "new schema", "new component", "write a hook".

Performance checklist for backend services and database queries — parallel async, N+1 detection, single-pass pagination counts, index coverage, STABLE function reuse, column selection, EXPLAIN-evidence for rewrites, and race-condition handling. Use when writing or reviewing any backend endpoint, service, or DB query (SQL, Drizzle, Prisma, Convex), when adding a WHERE clause or index, or when a perf rewrite/batching/join restructure is proposed. Triggers on "optimize this query", "endpoint is slow", "add an index", "N+1", "race condition".

Senior creative-director voice for design and branding work. Routes to one of 7 modes — Brief Simplify, Logo Concepts, Layout Compose, Typography System, Color Palette, Design Critique, Brand Identity System — based on the user's request. Use when the user says "design director", "simplify this brief", "logo concepts", "logo ideas", "layout ideas", "composition", "typography system", "font pairing", "type hierarchy", "color palette", "brand colors", "critique this design", "review this design", "design feedback", "is this design good", "brand identity", "visual identity system", "brand system", or asks for any design/branding direction. Accepts pasted images (Claude vision), text, or Figma URLs as input. Suggests downstream figma-* skills for implementation but never auto-chains.

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".

Run code review and coderabbit review in parallel on changed code. Use when user says "review", "code review", "run review", "check my code", "review changes", or after completing a task/feature. Also trigger proactively after significant code changes.

Deep project discovery and architecture planning. Use when starting a new project, when user says "new project", "let's build", or asks for help architecting/planning a non-trivial application.