catcatcat

Structured dissent skill. Takes a decision, plan, approach, or piece of work and pressure-tests it through adversarial analysis — not generic critique, but a concrete alternative argued at full strength. Articulates the case FOR the current approach, then spawns an adversary subagent to steel-man the best alternative, then synthesizes a decision brief with a mandatory verdict. Use ONLY when the user explicitly invokes "/adversary" — with or without additional context. Do NOT trigger on phrases like "what do you think" or "any concerns" without the /adversary command.

Staged build methodology for architecting and building software projects. Drives the full lifecycle: technical spike, design notes, build planning, and stage-by-stage execution with professional-grade testing. Use when: (1) user says "/architect", "architect this", "new project", "start a project", "build this from scratch", (2) user wants to plan and build a software project from scratch using a structured methodology, (3) user wants to resume an in-progress project that uses the staged build methodology, (4) user references the build methodology or staged build process. NOT for: quick scripts, one-off fixes, or tasks that don't warrant a full project structure. If the task is small enough to build in one shot without planning, just build it directly — don't force the methodology.

URL (YouTube, article, X/Twitter thread, podcast) or file path to extract from

Meta-cognitive skill that shifts the model from compliant executor to critical expert advisor. Identifies the domain, adopts top-tier expert perspective, understands the PURPOSE behind the work, and produces ranked proposals to elevate the current task, project, or output. Use ONLY when the user explicitly invokes "/elevate" — with or without additional context. Do NOT trigger on phrases like "make this better" or "what am I missing" without the /elevate command. Works across all domains: UI/UX design, copywriting, architecture, project planning, code, strategy, branding, etc.

Design judgment skill. Evaluates interfaces through the lens of a design director with taste — not checklists, not rubrics, not scores. Reads the design intent, assesses whether the execution serves it, and gives an opinionated verdict with clear direction. Scales from a single element to an entire app. Always judgment, never fixing. Use ONLY when the user explicitly invokes "/eye" — with or without additional context. Do NOT trigger on generic phrases like "review the design" or "how does it look."

"setup" = install testing infra only. "check" = audit gaps and report, don't write tests. "full" or omitted = setup + audit + write + run + verify.

Autonomous deep-work agent loop for driving projects forward with strategic intelligence. Each invocation performs one intelligent iteration: reads project state, assesses what's needed, applies critical expert thinking, does research when uncertain, executes one focused unit of work, and persists everything to notebook + PROJECT_STATE.md. Automatically sets up a /loop cron on first run (default 10m). Includes a three-tier confidence gate (GREEN/YELLOW/RED) that prevents the agent from acting when it should pause for human input. Use ONLY when the user explicitly invokes the command: "/liquid-cat-physics" or "lcp". Do NOT trigger on general phrases like "autonomous build", "deep work loop", or "drive this project forward" — those are not invocations of this skill. NOT for: one-off tasks, quick fixes, tasks without a project structure, or projects without clear goals. If the task is small enough to build in one shot without planning, just build it directly — don't force the methodology.

Context handoff for fresh sessions. Saves unsaved session knowledge to notebook, then produces a short orientation block to paste into a new chat. Use when: user says "/memento", "hand off context", "prepare for fresh chat", "save and brief", or before clearing context. NOT for: recovering context (use /notebook recover), saving a single note (use /notebook save), or replacing notebook.

Project notes system for preventing AI context loss and reasoning loops. Initialize notes, save notes mid-flow, recover context, migrate messy notes, and resolve lessons. Subcommands: /notebook, /notebook save, /notebook recover, /notebook migrate.

MUST invoke when user types "/prodev" or "prodev". This is a slash command — always use the Skill tool to execute it, never skip or summarize. Pro mode — engineering standard enforcement. Forces Claude to operate like a senior engineer who maintains what they ship. Prevents common AI coding failures: bandaid fixes, stale API knowledge, stubs, sycophantic agreement with bad approaches, and building on unverified assumptions. Loads a full field manual covering root cause discipline, layer awareness, research protocol, current-gen patterns, blast radius thinking, caller-first API design, and self-review.

Pre-build decision intelligence for any project. Produces a single opinionated decision brief — the output a senior consultant would give in week 1 before scope, stack, or approach are locked. Surfaces the traps that will cost you AND the edges that will make this an outlier — both calibrated to the project's actual scope (prototype vs production vs enterprise). Not a research report. A ranked decision brief. Use when: (1) user says "/recon", "what should I watch out for building X", "research best practices for X", "where's the edge building X", "what do good [X] projects do", "common mistakes with X", (2) user is about to start a new project and hasn't done research yet, (3) user describes a project idea and asks for advice before building, (4) user mentions a stack or category they haven't worked with before, (5) mid-build when hitting an area of genuine uncertainty about the domain. NOT for: debugging a specific bug (just debug it), researching a single library's API (just read docs), extracting knowledge from a specific URL (that's /eat), post-build optimization (that's /fortify), architecture planning (that's /architect).

A directory path, file path, or natural language scope like "auth layer" or "API routes". Omit to vet the entire project.

X/Twitter content intelligence with a social graph. Scout for reply opportunities, pulse-check topic lanes, track competitors, mirror your own accounts, prospect for clients. Graph-first design — builds around the accounts and topics YOU care about, not generic search. Use when: (1) user says "xray", "x research", "search x", "what's happening on x", "scout", "pulse", "track", "mirror", "prospect", "/xray", (2) user needs real-time X discourse on a topic, (3) user wants engagement opportunities or content ideas, (4) user wants competitive intel on tracked accounts. NOT for: posting tweets, account management, or analytics dashboards.