skills

Router for Ben's frontend-skills harness and every local skill.

Evidence-backed Go review for bounds, APIs, concurrency, errors, security, tests, rollout, and controllers. Use when reviewing a Go diff, PR, or branch, or as the golang hat /review adds when the diff touches *.go, go.mod, or backend protos.

Evidence-backed Go conventions for workload bounds, error boundaries, proto/API implementation, Temporal, testing, rollout, and controllers. Use when writing or changing Go services, handlers, workflows, controllers, or Go tests.

Evidence-backed Go review for bounds, APIs, concurrency, errors, security, tests, rollout, and controllers. Use when reviewing a Go diff, PR, or branch, or as the golang hat /review adds when the diff touches *.go, go.mod, or backend protos.

Evidence-backed Go conventions for workload bounds, error boundaries, proto/API implementation, Temporal, testing, rollout, and controllers. Use when writing or changing Go services, handlers, workflows, controllers, or Go tests.

Reviews a diff with a 7-hat parallel panel -- product/spec, standards, complexity, adversarial, resilience, visual/design, test/perf -- plus a mandatory cross-family GPT hat. Use when reviewing a branch, PR, WIP, or a release audit (deep mode).

Bootstrap the frontend stack -- toolchain, lint, quality gates, React/router/data enforcement, CI -- from lazy per-tool profiles.

Ship what built. Run verify -> self-review -> simplify -> deslop -> commit-push-pr -> monitor CI -> fix -> done. Use when implementation done, ready to launch.

Run a large rewrite, port, or migration using the Bun-in-Rust playbook -- baselines, mechanical-first translation, adversarial diff review, fix-the-process. Use on "rewrite X in Y", "port", "migrate the codebase", "revamp".

Delegate work to GPT-5.6 via the codex CLI -- clear-spec implementation, independent review, computer use, investigation, or data analysis. Use for bulk mechanical work, token-hungry tasks, or a cross-model second opinion.

Use when frontend/React/TypeScript/UI work. Guides understand -> plan -> grill -> TDD -> /go (4b->5). Alias: /work.

Use when a frontier model (Fable, Opus, GPT-5.6) should delegate bounded research, coding, testing, or log reduction to cheaper models while keeping planning, synthesis, risk, and final review central. Owns usage-limit budgeting for agent waves.

Explore options, then interview the user relentlessly about a plan, decision, or idea, with a 3-hat plan gate and optional domain-doc capture. Use when brainstorming approaches, starting new features, thinking before coding, stress-testing reasoning, as the lifecycle phase 2b gate, or on any 'grill' trigger phrase.

Murphy-law hat panel for unhappy paths -- input, timing, system, state, UX-recovery hats probe a change in parallel. Use when forms, validation, async/data, mutations, cache, state machines, destructive actions, or loading/empty/error states change.

Use when long-running or parallel agent work must respect 5-hour and weekly usage limits by checking usage between waves, pausing near the cap, and resuming only when the window is clear.

Parallel executor for independent bulk work across worktree lanes.

React accessibility patterns -- ARIA labels, keyboard handlers, widget attributes, nested-pressable fixes, AXE testing. Use when building or reviewing components, dialogs, forms, keyboard navigation, or fixing a11y findings.

Use when asked to watch, audit, compare, or fix another agent's work from a Codex/Claude session, transcript, thread link, PR, branch, log, or summary. Reconstruct the request, verify evidence, report gaps, and fix only when authorized.

Designs Google AIP-style protobuf resource APIs. Use when changing resource messages, standard-method RPCs, {resource}_id, parent wiring, CRUD/LRO shape, pagination, filtering, FieldMask, etag, or singleton resources.

Router for Ben's frontend-skills harness and every local skill.

Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs deep-module vocabulary.

Generate Codex hooks.json and AGENTS.md parity surfaces from the Claude hook manifest.

Delegate work to GPT-5.6 via the codex CLI -- clear-spec implementation, independent review, computer use, investigation, or data analysis. Use for bulk mechanical work, token-hungry tasks, or a cross-model second opinion.

Analyze changes, create categorized conventional commits, push, and open a PR. Use when the user asks to commit, commit and push, or open a PR. With --no-pr (or when the user asks for commit+push only), stop after Phase 4.

ConnectRPC + connect-query + Protobuf v2 patterns -- schema-first create(), typed transports, invalidation discipline. Use when writing API calls, mutations, query hooks, or proto-generated client code.

Write the least code that works and remove unjustified surface area. Use before commit/push/PR/merge, when a change feels overbuilt, or on ponytail, lazy mode, YAGNI, simplest solution, do less, bloat complaints, or debt-ledger asks.

Use when frontend/React/TypeScript/UI work. Guides understand -> plan -> grill -> TDD -> /go (4b->5). Alias: /work.

Diagnosis loop for hard bugs and performance regressions. Use when user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.

Build and sharpen a project's domain model. Use when the user wants to pin down domain terminology or a ubiquitous language, record an architectural decision, or when another skill needs to maintain the domain model.

Playwright + Testcontainers + axe-core E2E patterns for forms, tables, and workflows. Use when writing or fixing e2e specs, fixtures, browser tests, or debugging flaky Playwright runs.

Use when a frontier model (Fable, Opus, GPT-5.6) should delegate bounded research, coding, testing, or log reduction to cheaper models while keeping planning, synthesis, risk, and final review central. Owns usage-limit budgeting for agent waves.

Extend the frontend-skills hook harness: add rules, tune severity tiers, view analytics, debug non-firing hooks.

Bootstrap the frontend stack -- toolchain, lint, quality gates, React/router/data enforcement, CI -- from lazy per-tool profiles.

Ship what built. Run verify -> self-review -> simplify -> deslop -> commit-push-pr -> monitor CI -> fix -> done. Use when implementation done, ready to launch.

Explore options, then interview the user relentlessly about a plan, decision, or idea, with a 3-hat plan gate and optional domain-doc capture. Use when brainstorming approaches, starting new features, thinking before coding, stress-testing reasoning, as the lifecycle phase 2b gate, or on any 'grill' trigger phrase.

Compact the current session into a handoff document for another agent or fresh session.

Analyze hook effectiveness + session retro from collected metrics. Use when auditing hooks, inspecting hook latency, top violations, zero-fire rules, manifest drift, session trends, or running a retro across recent sessions.

Surveys a codebase as a senior advisor and writes implementation plans. Use when asked to audit code, find improvements, suggest roadmap direction, create handoff plans, review plans, dispatch execution, or reconcile backlog.

Clean noisy PR history and add reviewer guidance without changing code behavior.

Use when comparing, cross-reviewing, merging, judging, choosing, or arbitrating competing plans from agents, transcripts, visual-plan links, PR descriptions, files, or pasted strategies.