Coding & Development

Turns a NestJS 12 application into a Model Context Protocol server with @nestjs/mcp — McpModule.forRoot/forRootAsync, @Tool with Zod or Standard Schema inputSchema/outputSchema and annotations, @Resource and @ResourceTemplate, @Prompt, McpContext (authInfo, signal, reportProgress, inputRequired elicitation), guards, interceptors, pipes and filters with McpExecutionContext, tool errors vs JSON-RPC errors, OAuth 2.1 resource-server auth with McpTokenVerifier, scopes and @nestjs/authentication inte

Builds ready-to-use iPhone home screen widgets for the Scriptable app from one sentence. Use when someone asks for an iPhone widget or a Scriptable script.

Audit a codebase through every lens its stack calls for, verify each finding, cluster them by root cause, deliver the fixes as a ticketed map, and untangle the structure.

Write the simplest, most human-readable code that produces the required output, and strip out complicated logic, abstractions and extras that do not change that output. Use this whenever you write, extend or refactor code in any language, even if the user only says "build", "implement" or "add", and especially when they ask for code that is simple, readable, clean, maintainable, easy to understand, or not over-engineered, or complain that AI code is too complicated. Also use it when a request sa

Designs and reviews APIs, interfaces, and contracts for correctness, consistency, security, and evolvability. Use when defining endpoints, request and response schemas, errors, auth, pagination, versioning, or integration boundaries. Trigger on "API", "endpoint", "эндпоинт", "contract", "interface", "спроектируй API", "контракт".

Act as a software architect. Define or audit the technology stack and record

Build, check, and edit Fritzing sketches (.fzz and .fz files) with code. Makes the breadboard, schematic, and PCB views from one net list, adds custom parts (.fzp and SVG files), and proves the file loads by exporting SVG files and by opening it in the real Fritzing window. Use this skill whenever the user mentions Fritzing, a .fzz, .fz, .fzp, or .fzpz file, or asks to wire up, lay out, or draw a circuit on a breadboard, as a schematic, or as a PCB in Fritzing. Also use it when a Fritzing part d

Run an end-to-end game production pipeline with hard stage gates — idea to brief, GDD, tech plan, art and audio specs, placeholder assets, a compilable project from templates, build verification, and release plan. Use when the user wants the whole flow from an idea to a shippable build, not just one document.

Run a release preflight on a Flutter project before building for Google Play or the App Store. Use when the user asks "is my app ready to release", "check before I publish", "release checklist", "why was my app rejected", or before running flutter build appbundle / ipa.

Guide a new or substantially extended small public-facing Web project from clarified requirements through incremental implementation and verification. Use when the user wants a structured AI coding workflow that keeps product decisions, code changes, and acceptance results aligned.

Audit Next.js and Supabase SaaS projects before launch or after AI-generated changes. Use for authentication or tenant-isolation reviews, Supabase RLS and storage policies, accidental client-side secret exposure, checkout entitlement checks, and signup/login/checkout regression planning. Produces evidence-linked findings, scoped fixes, and verification results rather than claiming that a static scan certifies security.

Add email-and-password auth to a project (sign up, sign in, sign out, sessions, roles, change password, forgot/reset password) from a tested template instead of writing it from scratch. Use when the user asks to add auth, login, sign-up, accounts, sessions, protected routes or password reset to a FastAPI, Next.js or Node backend.

Diagnose Stripe webhook signature failures, duplicate fulfilment, missing payment credits, and out-of-order payment or refund events. Use when a user says a Stripe payment succeeded but their order is unpaid, a webhook credits twice, signature verification fails, or a refund ledger drifts. Produces evidence-based diagnosis, regression tests, and a repair plan; includes an offline verifier and durable test-mode reference ledger.

Turns an approved PRD into a tracked set of task files. Use when a PRD exists at docs/prd/ and needs executable tasks before implementation. Use when the user says "get tasks", "breakdown PRD", "pecah jadi task", or "bikin task list". Produces one detailed file per task with acceptance criteria, five-way verification, blocking edges, rollback, a plan document indexing them, and a parallelization classification.

Audit the full PR for merge readiness across correctness, maintainability and complexity, and test quality, then publish one actionable comment. Covers the contribution, not just existing feedback. Use when the user asks for a PR audit or whether a PR is ready to merge.

Runs a deep structured interview and produces a PRD in a single pass. Use when starting a new application, system, or feature and no requirements document exists yet. Use when the user says "get prd", "bikin PRD", "mau bikin sistem baru", "grill me", or asks to be interviewed toward a spec. Combines design-tree interviewing, intent extraction, capability mapping, and PRD authoring.

Builds a reusable skill, agent or main agent file (CLAUDE.md) from terms a person agrees, and keeps those terms as a contract its file is generated from. Use when someone wants a new skill or agent, wants to turn a prompt they keep pasting into one, or wants to change or write the contract for one. Not for a one-off prompt, for running a skill or agent that already exists, or for a legal or business contract.

Design, write, modify, review, or optimize MySQL data models, table structures, SQL, indexes, transactions, and concurrency control. Use when the task or project context establishes MySQL as the target database; do not select MySQL for generic database tasks or apply this skill to other databases.

Use when the engram memory MCP tools are available (engram_memory_search, engram_memory_get, engram_memory_capture, engram_memory_request_admission, engram_memory_withdraw, engram_memory_forget) and you need to recall durable project knowledge, record a cited durable decision or verified fact, request its admission, withdraw your own pending candidate, or forget an admitted record, including inside an h2a session. Do not trigger for scratch state, chat transcripts, tool-output dumps, or...

Engineering discipline for working in java. Covers: a speculated path deoptimises when the input changes; a view is not a copy; a wrapper type is not free here; close what you open; design for inheritance or forbid it; equality is one contract; exceptions name what failed; identity is not equality for boxes; no reference to internals escapes; null is not a value; publish safely or not at all; seal the alternatives; serializable is a second constructor

Audit a whole repository — not a diff — against the rules the other skills already apply to a change: the security checklist, code provenance and licensing, the family's code-quality rules, and whether the repository is ready for agent-driven development (CLAUDE.md, AGENTS.md, an ignored profile, a detectable test command). Asks you in the chat, in one batched round, the decisions that block a fix — which licence, whether a copyleft dependency may stay, where a copied block came from — and...

Engineering discipline that applies to every project and language. Covers: guarantee needs a reader; verdict survives the channel; check the claim you inherit; doc currency; measure the claim not a subset; no silent spend; prefer by construction; repair the lying artefact; signal needs a consequence; verify through production path; detector excludes own definitions; an ambiguous write target is a question; anchor outside the checked partys reach; delegate a fan out; diff against the known...

Engineering discipline for working in low-latency. Covers: attack the design in a second pass; a measurement matches the regime it reports; a queue without a bound has no overload behaviour; a structure keeps the regime it was proved under; allocated is not resident; answer the requirement at its layer; no allocation on the hot path; no coordinated omission; no false sharing on a hot line; no retry loop on a contended path; no stall inside a publication window; no syscall on a bounded path;...

Engineering discipline for working in rust. Covers: design types first; borrow in signatures; module visibility is deliberate; must use on consequential returns; newtype liberally; parse wide then range check; seal closed trait sets; typestate builder for required fields; typestate for protocols; verify the abstraction compiled away; async all the way; errors name what failed; justify every clone; no let underscore on a local; parse dont validate; private fields only; xplat fixtures; no...