Narraitor: A world-agnostic narrative-driven RPG framework using AI
Skill Claude CodeCodex
The evidence bar for AI narrative quality in Narraitor. Use when judging whether a prompt or generation change "works", when someone says "the output reads better", "the AI handles it", "works on my world", or when evaluating narrative coherence, choices, endings, or any Gemini-generated behavior. Encodes why one good…
Skill Claude CodeCodex
The load-bearing architecture invariants of Narraitor - domain boundaries, Zustand store contracts, the event bus, persistence rules, Next.js 15 server/client boundaries, and what may never be bypassed. Use before changing a store's shape, adding cross-domain imports, touching src/lib/api or src/lib/ai seams, moving…
Skill Claude CodeCodex
Automatically invoke when creating new components, stores, API routes, or planning a feature in Narraitor. Triggers on: new file creation under src/, architecture questions, feature design discussions, or any structural decision. Applies domain boundaries, naming conventions, state management patterns, and AI…
Skill Claude CodeCodex
Recreate, run, and verify the Narraitor environment - install, dev server, Storybook, the full quality gate, and how to classify failures as environment vs data vs AI-provider vs code. Use when setting up from scratch, when "npm test / build / lint fails", when the dev server won't start or the port is taken, when…
Skill Claude CodeCodex
How changes are classified, gated, evidenced, and shipped in Narraitor - and what may never be claimed without proof. Use before declaring anything "done", "fixed", "validated", or "verified", before re-baselining any gate (visual snapshots, dep-cruiser, skott, ds-canon), when correcting an earlier wrong claim, or…
Skill Claude CodeCodex
Symptom-to-triage runbook for Narraitor failures - blank narrative, AI route 500/timeouts, hydration mismatch, state not persisting after reload, works-in-Storybook-but-broken-in-app, theming drift, flaky visual specs, mystery tsc errors. Use whenever something is broken and the cause isn't already proven, or when…
Skill Claude CodeCodex
How to MEASURE instead of eyeball in Narraitor - curl smokes for the AI/API routes, inspecting persisted Zustand state in IndexedDB, DOM/computed-style checks for layout and theming, visual tooling, and where reports land. Use when you need proof of runtime behavior - "is the route up", "what's actually in the store"…
Skill Claude CodeCodex
Maintaining Narraitor's docs of record - status language, marking stale claims, append-only corrections, doc templates, and the project voice for all generated prose (PRs, issues, comments, code comments, user-facing copy, docs). Use when WRITING or FIXING any documentation or prose, when marking/correcting a doc that…
Skill Claude CodeCodex
Applied domain knowledge a mid-level engineer lacks for Narraitor - the world/character/narrative/session data model, Zustand store and persistence conventions, Next.js 15 App Router pitfalls, streaming/latency realities of Gemini, and the single-design-system (ds3) theming model. Use when reasoning about how domains…
Skill Claude CodeCodex
Narraitor's memory of painful failures, reverts, dead ends, and deliberately settled decisions. Use BEFORE proposing to remove/replace something that looks wrong or overengineered (the event bus, the react-joyride pin, split store files), before re-attempting an idea that may have been tried, when a doc or memory…
Skill Claude CodeCodex
How a hunch becomes a shipped, held, or retired change in Narraitor - issue capture, acceptance criteria, experiment flagging for AI-behavior changes, the gate sequence, and ship/hold memos. Use when starting any feature or improvement idea, when someone says "let's try", "quick experiment", "I have an idea for the…
Skill Claude CodeCodex
Executable campaign for Narraitor's hardest live problem - closing the v1.0 release gate, whose un-automatable core is validating the live AI play loop (real Gemini, multi-turn, consequences, endings) that every automated suite deliberately gates off. Use when asked to "get v1.0 out", "close the release gate"…
Skill Claude CodeCodex
Run a batch of Narraitor issues as parallel worktree lanes, with a Fable orchestrator picking the batch and delegating each issue to a subagent whose model matches the issue's model-power label. Use when asked to "work the top N issues in parallel", "fan out the backlog", "kick off a parallel batch", "orchestrate…
Skill Claude CodeCodex
Automatically run after any code change in this repo (components, stores, API routes, CSS, tests, types). Triggers on: new files, refactors, feature additions, or any edit touching src/. Reviews changes for alignment with established project patterns — design tokens, state conventions, error handling, domain…
Skill Claude CodeCodex
Drive real multi-turn game sessions in a browser against live Gemini and score whether the story is actually engaging. Use when asking "is this fun to play", "does the narrative hold up over 30 turns", "why does the story get boring", when planning a release playtest, or when a change to prompts or narrative logic…
Skill Claude CodeCodex
Narraitor's open frontier problems and how to attack them in THIS repo - long-arc story memory and coherence, consequence tracking, progressive streaming, replayability, and automated prompt evaluation. Use when asked "what's next", "what would make this state of the art", when scoping ambitious AI-narrative work, or…
Skill Claude CodeCodex
The governed lifecycle for AI prompt templates in Narraitor - registry, input contract, context-leakage gate, evaluation, regression, cost, and integration gates. Use when adding or editing any prompt template, when tempted to inline a prompt string into a generator or route, when reviewing a prompt change, or when…
Skill Claude CodeCodex
Fast orientation map for a zero-context session in the Narraitor repo. Use FIRST when starting work here, when asked "where does X live", "how is this repo organized", "what's the stack", "where do I look for the stores / AI code / themes / tests", or to check WHICH docs are trustworthy before acting on one. (Asking…
Skill Claude CodeCodex
Isolation-vs-integration parity reasoning and the component promotion ladder for Narraitor. Use when something "works in Storybook" but not the app (or vice versa), before calling any component done/integrated, when writing stories with withStores or MSW, or when deciding whether Storybook-green is enough evidence.…
Skill Claude CodeCodex
What counts as evidence in Narraitor - the test tiers (unit, Storybook, visual regression, e2e, manual QA walkthrough), what each can and cannot prove, acceptance thresholds, how to add tests, and how to read failures. Use when deciding which tests a change needs, interpreting a red or suspiciously green run, planning…
Skill Claude CodeCodex
Use when reviewing Narraitor code or PR changes for bugs, regressions, missing tests, and merge readiness.
Skill Claude CodeCodex
Port inline styles from a reference/demo component to production CSS in the Narraitor design system. Rigid 6-phase process that diffs inline styles against production stylesheets, ports gaps using design tokens, and removes inline styles from production components. Use when porting design system styles from any…