boilerplate-vite-supabase-mui-cursor

Bundles all uncommitted work from multiple Cursor agent threads on the same branch and checkout into one finish commit, then pushes. Use when the user runs /bundle-ship, /finish-all, or asks to finish and push everything across threads on one checkout.

Interview the user until scope edges and interaction boundaries are aligned and the decision tree is resolved. Focuses on where the feature stops, what it will NOT do or touch, and how it meets other functionality. Chat-only alignment — authors no artifact. Use to stress-test scope/boundaries before feature or plan, or when the user says "grill me".

Picks the next action and skill, then invokes it (read that SKILL.md and execute the workflow in the same turn). Mid-thread: continue the logical next step on active work. Idle or new: fall back to gates and app-tasks.json backlog. Maps situations to project, user, and plugin skills; resolves overlaps. Product context SSOT: documentation/DOC_APP_VISION.md. Use when the user types /router, asks which skill to use, starts ambiguous work, or routes slash commands to SKILL.md files. Bare /router is next-action dispatch (not finish).

Challenges and simplifies a concrete feature to reduce steps, clicks, and cognitive load (flow and code modes by default). Use when the user targets a workflow and asks whether it is overbuilt, or runs /challenge.

Systematically audit the codebase for consolidation, abstraction, and redundancy removal opportunities across features. Use when the user asks to find duplication, consolidate code, audit for redundancy, reduce repetition, or identify shared patterns that should be unified.

Structured scientific debugging for runtime bugs, regressions, flaky behavior, frontend/backend failures, Supabase/RLS/Auth/Storage issues, and user-reported errors. Use when the user invokes /debug, asks to debug, reports an error, or describes inconsistent behavior.

Runs systematic new-feature engineering with mandatory decision stops; produces product/requirements artifacts, not DEVELOPMENT_PLAN.md (plan is SSOT for execution plans). Use for new feature requests before planning. Not implement, validate, or finish.

Completes implementation wrap-up: concurrent-agent smoke checks, version/changelog when required, commit, and coordination before push. Use when closing a job, after implement/validate, or when the user runs /finish. Not push-only (use push) and not planning (use plan).

Debug by pre-registered hypothesis instead of guess-and-patch. Before touching code, the agent states what it believes the root cause is, predicts what a successful fix proves, and — critically — pre-registers what a FAILED fix would teach and how that narrows the space of remaining causes. Use this whenever debugging a bug, error, test failure, or "why isn't this working" issue, OR whenever the user explicitly invokes hypothesis mode (e.g. "use the hypothesis skill", "debug this with hypotheses", "hypothesis debugging"). Especially use it for non-obvious bugs where the cause is unclear, intermittent failures, or bugs that have already survived one or more naive fix attempts. The goal is to corner the cause across iterations, not to fix in one shot.

Executes DEVELOPMENT_PLAN.md phase by phase with gates, progress updates, and repo-rule compliance. Use when the user asks to implement a plan or continue a job. Not planning (plan), changelog (finish), or read-only audits (validate).

Analyzes code for optimization across design, approach, efficiency, and complexity so fixes happen at the right level. Use when the user asks to optimize, reduce complexity, refactor thoughtfully, or assess whether code should be rewritten vs tweaked.

Creates DEVELOPMENT_PLAN.md with repo-rule compliance researched first (file placement, architecture, patterns). Use for new features, M/L work, or /plan. Not XS/S quick changes (quick-piv), product-only requirements (feature), or changelog updates (finish).

Pushes previously finished commits to remote only; never stages, commits, or amends. Use when finish has already run, the working tree is clean, and local commits exist. If uncommitted changes remain, stop and run finish first.

Orchestrates compressed Plan → Implement → Validate for small XS/S scoped changes in one session without a full DEVELOPMENT_PLAN.md. Use for focused fixes and tweaks. Not M/L work, migrations, or when plan review or pattern-review gates are required (use full plan cycle).

Scores a TypeScript + MUI component against a structured quality rubric (0–5 per item). Use when reviewing a specific component for UX, architecture, and maintainability. Not repo-rule audits (validate) or development-plan critique (review-dev-plan).

Human onboarding: README Quick Start, dev task backlog (/tasks), gates, verification. Requires filling documentation/DOC_APP_VISION.md (problem, persona, app role) before fork/clone unless user defers. Use for first-time setup and /start-style requests.

Validates a plan or an implementation via parallel read-only rule-shaped audits, then synthesizes findings before any fixes. Auto-selects depth: plan review, full implementation review (rules + tooling + plan-compliance), or a lighter pre-merge architecture gate. Use when validating a plan/implementation, before merge/finish, or after large refactors.

Audit and improve the .agents/skills library so the skills form a symbiotic whole: clear separation of concerns, single source of truth (SSOT), and no routing or instruction conflicts. Runs parallel Task subagents per lens (inventory, separation/overlap, SSOT/duplication, conflicts, composition/handoffs) and ends with a critical no-information-loss pass that must pass before any edit is kept. Use when the user asks to harmonize, consolidate, deduplicate, or de-conflict skills, fix router/skill overlaps, or improve the skill library as a whole.

Loads project context at session start (structure, governance, active jobs, enforced boundaries). Use for a new chat, ambiguous tasks, or before larger refactors. Not human onboarding (start) or pre-merge gates (validate gate depth).

Stack-native React performance patterns for this Vite SPA (React 19, MUI, TanStack Query, Cloudflare Workers). Use when optimizing bundles, waterfalls, re-renders, or client fetch behavior — not for repo-wide refactors (optimize2) or Lighthouse audits (web-perf).

Inspect an Airtable base in two phases: (1) schema — tables, field ids/types, links via Meta API and repo SSOT, no row data; (2) sample — a few rows to see cell shapes (attachments, rich text, URLs, linked-record ids) with fld… keys. Always schema first, samples second. Safe defaults: low maxRecords, optional fields allowlist, redact PII. Triggers: Meta API, field id, linked tables, schema drift, sample rows, cell shape, returnFieldsByFieldId, wire JSON vs app normalization.

Turns recent code changes, failures, and conversation mistakes into durable guidance. Use when the user runs /learn, asks for a retrospective or to sharpen rules, or after resolving a multi-turn struggle (e.g. 3+ failed attempts, critical CI fix).

Writes cross-repo adoption guides (<SLUG>_ADOPTION_GUIDE.md) from an implemented pattern in the source repo. Layer-agnostic porting steps, what-is/is-not, reference file map, and target-repo checklist. Use when the user wants an adoption guide, playbook for other repos, or to document how to replicate a capability outside this codebase. Does not require the learn skill.

Runs six parallel Task subagents to critique a development plan (feedback only). Use when checking whether a plan is sound before implementation; no code or plan edits unless the user asks afterward.

Grade or improve an attached rule/command file. Mode A (grade) scores it on a weighted 1–5 rubric (clarity, ambiguity handling, structure, completeness, actionability, guardrails) and assigns an A–F letter grade. Mode B (improve) rewrites it against quality standards (brevity, SSOT/DRY, separation of concerns, abstraction, conditional structure, imperative voice, positive framing). Use when the user wants a rule/command scored, critiqued, or tightened.

PROACTIVE: Compare a development plan or implementation proposal to industry standards and common best practice before application code — without waiting for the user to ask. The agent selects relevant aspects (UX, APIs, identity, async, coupling, etc.) per change. Use at planning (especially M/L), feature architecture, router intake on novel behavior, review-dev-plan industry lens, or when the user asks for best practice / industry standard check. Not repo rule compliance (use validate for that).