Claude Skills by The-Utopia-Studio

This skill should be used when the user wants to push code to Railway, says "railway up", "deploy", "deploy to railway", "ship", or "push". For initial setup or creating services, use new skill. For Docker images, use environment skill.

This skill should be used when the user wants to manage Railway deployments, view logs, or debug issues. Covers deployment lifecycle (remove, stop, redeploy, restart), deployment visibility (list, status, history), and troubleshooting (logs, errors, failures, crashes, why deploy failed). NOT for deleting services - use environment skill with isDeleted for that.

This skill should be used when the user wants to add a domain, generate a railway domain, check current domains, get the URL for a service, or remove a domain.

This skill should be used when the user asks "what's the config", "show me the configuration", "what variables are set", "environment config", "service config", "railway config", or wants to add/set/delete variables, change build/deploy settings, scale replicas, connect repos, or delete services.

This skill should be used when the user asks about resource usage, CPU, memory, network, disk, or service performance. Covers questions like "how much memory is my service using" or "is my service slow".

This skill should be used when the user says "setup", "deploy to railway", "initialize", "create project", "create service", or wants to deploy from GitHub. Handles initial setup AND adding services to existing projects. For databases, use the database skill instead.

This skill should be used when the user wants to list all projects, switch projects, rename a project, enable/disable PR deploys, make a project public/private, or modify project settings.

This skill should be used when the user asks about Railway features, how Railway works, or shares a docs.railway.com URL. Fetches up-to-date Railway docs to answer accurately.

This skill should be used when the user asks about service status, wants to rename a service, change service icons, link services, or create services with Docker images. For creating services with local code, prefer the `new` skill. For GitHub repo sources, use `new` skill to create empty service then `environment` skill to configure source.

This skill should be used when the user asks "railway status", "is it running", "what's deployed", "deployment status", or about uptime. NOT for variables ("what variables", "env vars", "add variable") or configuration queries - use environment skill for those.

This skill should be used when the user wants to add a service from a template, find templates for a specific use case, or deploy tools like Ghost, Strapi, n8n, Minio, Uptime Kuma, etc. For databases (Postgres, Redis, MySQL, MongoDB), prefer the database skill.

Stands up the post-launch loop for a shipped AI product — ship behind sign-off → observe tweak-time (how much the customer edits before accepting) → learn by re-running the WHOLE eval set on every model release → refine by promoting autonomy only as a scored eval result, never by feel. Fires on "what happens after launch", "improve the product", "set up the loop", "a new model dropped, should we upgrade", "can we raise the autonomy / drop the sign-off yet". Returns a filled flywheel cadence d...

Scans a GitHub repository to understand its tech stack, hosting, monitoring, integrations, and current production readiness. Use when the user asks to "audit", "check", or "scan" their repo. Don't use for code review or bug detection.

Cleans up and structures a repository for production. Adds README files to every directory, modularizes code, removes dead files, and makes the repo LLM-smart. Use when the user asks to "clean up", "structure", "document", or "organize" their repo. Don't use for code review or bug fixing.

Audits repository security — hardcoded secrets, dependency vulnerabilities, environment variable management, and authentication patterns. Use when the user asks to "check security", "find secrets", "audit dependencies", or "secure my repo". Don't use for code review, deployment, or monitoring.

Converts any block of text into a conceptual sketch image. Takes raw text, extracts the central idea, generates an image prompt, and produces the final image using ComfyUI + Flux.1 Schnell locally. Use when the user pastes text, notes, or ideas and wants a visual concept. Trigger on "sketch this", "turn this into a sketch", "visualize this", "make a sketch", "image for this", "sketch prompt", or when the user drops text and wants it converted into a conceptual illustration.

Fires when a fellow wants a whole idea pressure-tested fast across five business questions — "is this worth doing", "pressure-test this", "poke holes in this", "stress test my idea", "so what?". Runs So what / Who pays / What's the wedge / What's the moat / What kills it, then returns a per-question verdict table and an overall Pass / Redesign / Kill, refusing to pass any question answered on opinion. Do NOT fire to score a problem on the 8-dimension gate (use problem-quality-scorecard), for ...

Build a synthetic user panel grounded in your YODA corpus and pre-test a concept, message, or flow against it — cheaply, before you spend real interview time. Returns a Synthetic Panel + Pre-Test Transcript carrying a lie ledger (what synthetic users systematically get wrong: they never pay, never no-show, over-rationalise, over-agree) and a kill / revise / promote-to-real-interview verdict. Fires on "test this before real interviews", "simulate a user", "dry-run the pitch", "pre-test the mes...

Shadow an expert doing real work and extract the judgment calls they make without noticing. Fires when a fellow says "interview an expert", "capture how they decide", "watch them work", "shadow an operator", "capture the judgment", or "encode when they override the rule". Runs a watch-30 / replay-20 / edge-cases-10 session and returns a Tacit-Knowledge Capture: the explicit procedure split from the tacit judgment, each fork tagged with the cue read and the exception that overrides it. NOT for...

Enforces a disciplined Red-Green-Refactor (TDD) workflow in TypeScript/Node.js. Use this whenever creating new features, fixing bugs, or migrating logic to ensure high-quality, verifiable implementations.

Turns a set of production traces into discovery interviews — recovering, for each behaviour pattern, the discovery question the user answered by acting instead of talking. Fires on "learn from usage", "mine the logs", "what are our production traces telling us", "turn usage data into discovery", "what are users saying in the logs". Outputs a trace-interview log: one card per behaviour pattern with the observed behaviour [Fact, 0.7], the discovery question it answers, the inferred JTBD/satisfa...

Architecture standard for building robust, type-safe TypeScript services using the "Spec and Handler" pattern. Use when building CLIs, libraries, or complex business logic.

Design engineering principles for making interfaces feel polished. Use when building UI components, reviewing frontend code, implementing animations, hover states, shadows, borders, typography, micro-interactions, enter/exit animations, or any visual detail work. Triggers on UI polish, design details, "make it feel better", "feels off", stagger animations, border radius, optical alignment, font smoothing, tabular numbers, image outlines, box shadows.

Ranks what a customer actually needs by onioning each stated need to a tech-invariant core, sorting the survivors with Kano (must / performance / delighter), and returning the one durable, under-served need that will still be a need after the technology turns over. Fires on "what do they really need", "find the unmet need", "which need should we prioritise", "what's the real need under this feature request", or when a fellow hands you a list of "needs" that are mostly features and wants the d...

Run a moderated usability session on a working prototype and return findings tied to the screen, not the user. Fires when a fellow says "run a usability test", "watch a user try it", "usability protocol", "can a first-time user complete X", or "see where they get stuck". Product runs, Engineering watches muted, Design notes; the task is framed as a goal not an instruction, the user is kept in user mode (no leading, no explaining, no "do you like it"), and every finding is scored on observed t...

Splits a buying situation into four distinct people — user (feels the pain, touches the product), buyer (owns the budget line), decision-maker (final signature / veto), champion (spends internal capital to push it) — each pinned to ONE title, ONE budget (or an explicit none), ONE why-now. Fires on "who's the actual buyer", "who decides", "map the stakeholders", "map the buying committee", "the user loves it but who signs the cheque", "who's our champion and who can veto us". Output is a fille...

Gate a built v1 to go / no-go on two recorded tests — a true stranger reaches first correct value in ≤60 seconds unaided, AND the learn-from-corrections mechanism is wired and proven to compound (the thing 95% of enterprise pilots never deliver). Output is a recorded-evidence go / no-go checklist. Fires on "are we ready to launch", "is v1 good enough", "launch bar", "go/no-go on v1", "should we ship this pilot". Not the ongoing post-launch loop or cadence (use refine-flywheel), not pilot pric...

Set the price and the charging metric for a validated product — anchor the number at 10–20% of the MEASURED incremental value, and charge on the outcome / asset / decision the customer gets, never on seats or tiers. Emits a value-priced model. Fires on "how should we price this", "what do we charge", "per seat or per outcome", "set the price", "what's the number". Not for brainstorming which revenue model to bet on across the product (that is `monetization-strategy`, exploratory; this commits...

React and Next.js performance optimization guidelines from Vercel Engineering. This skill should be used when writing, reviewing, or refactoring React/Next.js code to ensure optimal performance patterns. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.

Pressure-test ONE proposed wedge against five pass/fail thresholds (standalone value, forced cadence, travel, structural expansion, scoreable adoption) and return a 5-row verdict that calls it a real wedge only on 5/5. Fires on "is this a wedge", "is my wedge sharp enough", "what's my wedge", "the smallest thing that forces adoption", "will this actually get adopted". Not for defensibility over years (use moat-design-canvas), not for scoring the whole idea across eight dimensions (use problem...

Fires when a fellow wants a human to run the machine behind the curtain for a day to test usage — "fake the backend", "wizard of oz", "woz test", "human behind the curtain", "have a person answer the requests and see if they act on it". Output is a WoZ setup plus a result read-out that reports usage behaviour and an explicit "what this does NOT prove" section for feasibility, unit cost, and scale. NOT for choosing which probe to run (that is `probe-matrix`), NOT for testing whether a workflow...

Fires when a fellow needs to coordinate MORE THAN ONE agent to get the work done — "design the workflow", "orchestrate the agents", "set up the fleet", "how do the agents work together", "which agents do I need and when do they spawn", "multi-agent setup". Returns a workflow + fleet map: the multi-step orchestration, a spawn-by-rule fleet (researcher, analyst, prototyper, drafter, critic) each with an observable spawn trigger, and a Think/Build/Admin surface assignment (Claude.ai / Claude Cod...

Assemble a sourced proprietary-data corpus map — sort every candidate data source by one test (can a foundation model already reach it?) and keep only the workflow-trapped, permission-gated signal (shadow AI threads, operational logs, permits, vendor-locked sensor feeds, workflow exhaust) as the moat seed. Fires on "get our own data", "what data do we own", "where's our proprietary data", "build the moat's seed", "what data can't a competitor get". Not for drafting the contract that unlocks a...