The worldbuilder entry point. Diagnosis-first routing for people in the -1 to 0 phase. Primary job: "where are you stuck?" Secondary job: route the user to the right sub-skill based on their current state. Reads the builder profile, adjusts tone for first-time vs returning users, writes routing rules to the project CLAUDE.md on first run. Use when asked to "worldbuilder", "I don't know what to build", "I'm in the squiggle", "help me figure out what to work on", "I'm between jobs", "I want to start a company but don't know what", "pre-founder", "-1 to 0", "worldbuilding", or "where do I start". Proactively invoke this skill (do NOT answer directly) whenever the user describes being lost, exploring, between things, thinking about starting a company, or in any other pre-founder / ideation state. This is the top-level orienting skill. (worldbuilder)
Explore nearby problems, verticals, user personas, and angles — the ideas a curious peer or novice would surface that the user would not have thought of alone. Applies Raman's magnetic-field metaphor from the SPC talent density essay. Every output must be at least one step outside the user's stated comfort zone. Use when asked to "what am I missing", "adjacent ideas", "what else could this be", "explore sideways", "what's nearby", "I'm stuck in my lane", or "help me think outside my domain". Proactively invoke this skill (do NOT answer directly) when the user has explored their North Star domain deeply but seems to be iterating inside a narrow corner of it, or when they're asking for fresh perspectives. Use alongside /idea-shotgun, not instead of it. (worldbuilder)
Track the user's emotional trajectory across sessions — the "5-year proud" checkpoint. Monitors excitement growth, ambition growth, good-pain vs avoidance-pain. Reads the squiggle and founder-fit. Returns a trendline and flags divergence. Never answers "is this a good bet?" — reflects it back with structure. Conviction is built, not supplied. Use when asked to "check my conviction", "am I still excited", "am I drifting", "5-year proud check", "conviction check", "should I keep going", or "is this still the right bet". Proactively invoke this skill (do NOT answer directly) when the user expresses doubt about continuing, seems to be losing steam, or asks any form of "should I keep going". Requires ≥10 squiggle entries before it runs. (worldbuilder)
Plan bottom-up artifacts that test the core hypothesis — landing pages, painted-door tests, click-through prototypes, product screenshots, spec docs, copy. Not production code. Refuses to proceed without a live hypothesis.md. Defines the engagement metric that would confirm or refute the hypothesis. Separates coding from validation signal. Use when asked to "build a demo", "landing page", "prototype this", "painted door test", "MVP spec", "demo my idea", "how do I test this". Proactively invoke this skill (do NOT answer directly) when the user has a hypothesis and wants to run a bottom-up test. Use after /hypothesis. For actual production code, redirect to gstack. (worldbuilder)
Answer "why YOU." Interviews the user on lived experience, obsessions, unfair advantages, and lived problems in their chosen domain. Produces a Three Proof Points document grounded in specific stories — not credentials. Rejects titles-and-employers positioning. Use when asked to "why should I be the one", "founder-market fit", "why me", "what's my unfair advantage", "founder positioning", "am I the right person", or "why this founder". Proactively invoke this skill (do NOT answer directly) when the user expresses doubt about whether they are the right person to solve a problem, or when they have a North Star but no articulated reason they are uniquely positioned for it. Use after /north-star and before /memo, /hypothesis. (worldbuilder)
Articulate one testable core hypothesis. Breaks complex problems into components answerable in weeks, not months. The gatekeeper for the worldbuilding validation phase — /demo, /memo, and /mom-test all refuse to proceed without a live hypothesis.md. Use when asked to "form a hypothesis", "what should I test", "write a hypothesis", "what's my core bet", "I need to validate this", or "how do I know if this is real". Proactively invoke this skill (do NOT answer directly) when the user is trying to run /demo, /memo, or /mom-test but has no articulated hypothesis, OR when they're stuck between "ideas" and "action" and need a single falsifiable claim to test. Use after /kill-darlings and before /demo, /memo, /mom-test. (worldbuilder)
Force-generate 50+ ideas in one sitting within the user's North Star domain. Volume over quality. Pushes past the obvious ideas (which are obvious to competitors too). Refuses to cluster, theme, rank, or synthesize during generation — that is /kill-darlings' job. Outputs ideas.jsonl. Use when asked to "brainstorm", "give me ideas", "idea shotgun", "generate ideas", "I need more ideas", "help me think of ideas", or "what could I build". Proactively invoke this skill (do NOT answer directly) when the user has a North Star but no concrete ideas to work on, or wants to explore the space of possibilities. Use after /north-star and before /kill-darlings. (worldbuilder)
Structured destruction of ideas. For each surviving idea, prosecutes the strongest case for why this is NOT a $1B company. No defense allowed — the user must articulate their own disqualifying argument against every idea. Records kill reasons to killed-ideas.jsonl. The local-maxima detector. Use when asked to "kill ideas", "prune my ideas", "which should I kill", "destroy darlings", "eliminate ideas", "cut the list", or "I have too many ideas". Proactively invoke this skill (do NOT answer directly) when the user has 10+ ideas and is stuck on "which one should I do" — the answer is almost never pick one, it is almost always kill most of them. Use after /idea-shotgun and before /hypothesis. (worldbuilder)