haowjy/creative-writing-skills
18 skills18 installs4.3k starsGitHub
Install
npx skills add https://github.com/haowjy/creative-writing-skillsSkills in this repo
1BrainstormingThis brainstorming skill (Chapter Planning guidance) helps solo fiction and serial authors explore structure before drafting. The agent’s job is to mirror what you said about beats, scenes, pacing, and chapter openings or endings—not to upgrade a fuzzy plan into a three-act template or to write the fight you only named. It fits discovery in Idea when you are still shaping a chapter’s purpose, Validate when scoping whether a chapter earns its page count, and Build when you are maintaining a living chapter brief alongside the manuscript. Use it when you want a lightweight planning transcript the agent can respect during later writing skills, not when you already have a locked beat sheet and only need line edits.1installs2Cw Musecw-muse is the session lead for Haowjy’s creative-writing skill stack on Claude.ai, where subagents are unavailable so one conversation must cover every mode. Solo builders who write fiction, game narrative, or long-form creative IP install it at the start of a story session to get a partner that brainstorms directions, drafts prose, critiques honestly, and revises while maintaining the story knowledge base. The skill’s discipline is deliberate stance switching: brainstorming and drafting use immersion skills; critique uses an adversarial reader posture loaded from prose-critique and related craft files. It explicitly directs you to load brainstorming, writing-principles, prose-writing, scene-construction, and prose-critique as needed rather than improvising craft rules in chat. Prism places it at Idea discover because sessions begin with shaping what the story wants to be, but the same lead spans early exploration, iterative drafting (Build-like creative production), and revision passes (Ship-like quality tightening) for narrative deliverables.1installs3Grill With DocsGrill With Docs is an agent skill for solo and indie builders who have a plan or proposed direction but need it challenged against what is already documented. It interviews you relentlessly, resolving decision branches in dependency order so downstream choices are not built on vague language or unstated assumptions. Each exchange targets the next unresolved branch (or a tight cluster of related questions), includes a recommended answer, and waits for your reply—so you can run it in short sessions without losing thread. Terminology is checked against project vocabulary pages, CLAUDE.md conventions, active work notes, and prior decisions so new labels do not collide with established meaning. Invoke it when you want plan review that is sharper than a casual critique but lighter than a full formal spec process—especially before implementation plans, architecture choices, or narrative/product directions where words like “MVP,” “platform,” or domain terms are overloaded.1installs4Intent ModelingIntent Modeling is a journey-wide creative-writing and agent discipline skill that tells the model to understand author intent before obeying surface instructions. Solo builders hit failure modes when agents comply literally—“stop doing X” might mean rebalance toward Y—and when helpfulness overrides the actual ask. The skill instructs reading context, asking clarifying questions when ambiguous, and scanning for the same systematic misreading elsewhere once one slip is found. Use it whenever you delegate nuanced creative, product, or refactor work to an agent, not only at the start of a novel chapter. It pairs naturally with planning and brainstorming skills because approved intent should precede large implementation or rewrite passes. It does not replace explicit specs: when requirements are already signed off and unambiguous, extra intent parsing adds little. The payoff is fewer round trips and edits that match the outcome you wanted, not the cheapest interpretation of your last message.1installs5Kb ManagementKB Management is an agent skill for creative-writing projects that use a shared `kb/` directory as durable story memory. Solo builders and small writing teams install it when finalized chapters, character arcs, or world rules need to be captured as canon and synthesized into wiki pages every downstream agent can read. It explains how canon differs from living wiki reference, how style files anchor voice for writer and critic agents, how vocabulary and aliases stay consistent project-wide and per domain, and when to track multi-chapter issues elsewhere. The skill emphasizes one concept per document, splitting overloaded pages, and merging duplicate explanations so context windows stay focused. It is not a generic markdown linter—it is procedural knowledge for long-form narrative workflows where contradiction breaks reader trust. Use it whenever you promote draft facts to canon, refactor kb structure, or onboard new agents to an existing story bible.1installs6Llm Writingllm-writing is a journey-wide agent skill for solo builders who publish specs, landing copy, newsletters, and product docs with Claude Code or similar agents. It does not replace voice or strategy; it makes you notice when the model fills structure without purpose, summarizes with labels instead of mechanisms, states conclusions without evidence, smooths uncertainty, encodes fixes as brittle prohibitions, or restates the last paragraph as a fake transition. A dedicated section on conversational bleed catches documents written like chat replies—correcting misconceptions the reader never had or inviting them to “break something down” when no question was asked. The skill tells the agent to load intent-modeling first so revisions stay tied to audience and intent rather than generic “sounding professional.” Use it whenever a written artifact is meant for a reader who was not in your session, from validate landing pages through launch SEO posts and grow lifecycle emails.1installs7Project SetupProject Setup is an agent skill for solo fiction and serial authors who want Claude Code (or similar agents) to respect their voice, world, and folder habits from day one. Instead of repeating context in every chat, the skill runs a structured interview: project shape, progress, POV and timeline complexity, and where files already live. It gathers writing samples into kb/samples/, outlines a knowledge-base layout, and writes CLAUDE.md so every future creative-writing skill in the repo shares the same rules. When samples are sufficient, it can kick off style-analysis to seed kb/styles/. Use it once at project start or when migrating a messy draft tree into an agent-friendly structure—before heavy drafting, revision, or continuity checks.1installs8Prose CritiqueProse Critique is a reference-style agent skill for narrative fiction that treats praise without scrutiny as harmful. It instructs the agent to challenge the assumption that a draft already works, then go deep on a writer-specified focus area using linked resources for structure, character, voice, line-level prose, and continuity. Solo and indie builders who ship story-led products—serial fiction, interactive fiction, game narrative, or brand storytelling—can invoke it whenever they have a draft worth stress-testing, not when they only want validation. The workflow favors honest gaps (pacing, motivation, POV drift, show-vs-tell, timeline errors) over generic compliments. Because critique applies across ideation notes, validation prototypes, build-time manuscripts, and pre-launch polish, it behaves as journey-wide process knowledge anchored on the docs shelf for discoverability.1installs9Prose WritingProse-writing is a reference agent skill for narrative fiction at the sentence and paragraph level. Solo builders and indie authors who use Claude Code or similar agents to draft stories install it when they want immersion mechanics—psychic distance, free indirect discourse, and rhythm between close and far narration—rather than generic chat tone. The skill deliberately does not define project voice; you pass style files alongside it and read them before writing. It fits anywhere you revise prose: early idea sketches, validate-phase story prototypes, build-phase docs with narrative flair, or grow-phase serialized content and marketing fiction. Because it is craft reference rather than a one-shot generator, it works best when the agent iterates on scenes you are actively revising. Optimized for fiction and narrative content products, not API design or app scaffolding.1installs10Scene ConstructionScene Construction is a reference agent skill for creative writers and solo builders who ship content products—newsletters, fiction, games, or course scripts—and need scenes that work on the page. It explains how to enter scenes in medias res, run dialogue that does more than one job at a time, use subtext instead of on-the-nose exposition, differentiate voices, and pace beats through transitions. The skill is marked non–model-invocable reference material, so you pull it in when drafting or evaluating scenes rather than expecting automatic invocation. It intentionally defers sentence-level immersion to prose-writing and macro structure to story-architecture. Intermediate complexity: assumes you already have characters and plot pressure and want tighter scene mechanics for grow-phase content you publish repeatedly.1installs11Shared Daoshared-dao is a journey-wide creative-writing agent skill that establishes shared vocabulary between author intent and agent behavior. Solo novelists and serial fiction builders use it when magic systems get renamed, factions blur, or genre labels mean different things in different prompts—problems that corrupt drafts before structure is solid. The skill teaches where terms belong at the lowest shared scope: whole-project vocab in kb/vocab.md, domain-specific terms under kb/worldbuilding or factions, and provisional naming in work/ until promoted. It pushes aggressive scrutiny of important terms: reuse when names already fit, converge on one name per concept, and flag when only the author can judge. It is not a plot generator; it is the dao—the contract—that keeps retrieval and drafting aligned. Load it when establishing canon, resolving ambiguous names, or auditing consistency across kb/ before long agent-assisted writing sessions.1installs12Story Architecturestory-architecture is a generative-writing agent skill that teaches how to structure a narrative arc around a single dramatic question. It walks through hook, rising complications, a midpoint that reframes the conflict, a crisis requiring a costly choice, and a resolution that changes characters and tees up the next arc. Solo builders use it when shipping fiction, serialized content, interactive fiction, or game quests where sagging middles and flat stakes kill retention. The skill emphasizes logical escalation from character decisions rather than obstacle roulette, and it treats each arc’s answer as the seed of the following arc’s question. It is methodology, not prose generation: you get architectural vocabulary and pacing discipline your agent can apply to outlines, beat sheets, and season bibles before drafting.1installs13Story Contextstory-context is a reference agent skill that teaches how much creative-writing context to give spawned agents and which delivery mechanism to use. Solo builders running multi-agent fiction or long-form content pipelines install it to avoid the classic failure modes: writers who contradict established canon because they never saw the wiki, critics who miss continuity because history was stripped, and brainstormers who re-explore rejected angles. It complements spawn-command skills by explaining when stable files beat ephemeral session history, when to materialize decisions before handoff, and how to scope `-f` lists tightly. The skill is judgment-heavy rather than a single automation step, so it stays relevant across drafting, revision, and critique spawns. Use it whenever you delegate chapters, outlines, or editorial passes to child agents in Claude Code or similar environments.1installs14Style AnalysisStyle Analysis is a reference agent skill for creative-writing stacks that need reproducible voice control across writer and critic agents. It walks you through reading existing prose to discover which style dimensions vary independently—character voice, scene type, pacing, POV—and how those axes interact before you commit to a file layout. The skill emphasizes file splitting as a context engineering decision: every boundary should answer whether an agent would plausibly need one style chunk without another, which keeps orchestrator prompts lean when you pass individual files with `-f`. Solo builders shipping fiction pipelines, interactive story products, or branded long-form content with multiple agents install it when style drift between chapters or modes becomes expensive to fix manually. Because it is reference material rather than a one-shot generator, you use it while authoring or refactoring style libraries, not as a substitute for human taste on final copy. Pair it with writer and critic skills in the same repo so evaluations reference the same dimensional model you documented.1installs15Writing ArtifactsWriting Artifacts is a reference skill for solo authors and indie teams using Meridian-style multi-agent creative workflows. It answers a boring but failure-prone question: where does durable project knowledge live versus throwaway iteration files? Durable material belongs in the knowledge base directory governed by kb-conventions; active outlines, draft versions, critique rounds, and brainstorm captures live under a work directory scoped to the current item and archived when done. The skill stresses that humans, orchestrators, and workers share one tree—always re-read files before acting because the author may have edited between rounds. Use it at the start of any writing sprint, when wiring a new repo’s AGENTS.md, or when an agent is about to write a draft path that would clobber manual edits. It does not write prose; it prevents workspace chaos.1installs16Writing Issueswriting-issues is a reference agent skill for fiction and long-form creative writers who use AI critique partners during Build and Ship cycles. It teaches when to elevate a finding into a tracked issue: recurring tics, structural weaknesses that span scenes, or inconsistencies that deserve their own revision lane instead of disappearing after one report. Solo builders shipping newsletters, serialized fiction, or narrative-led products benefit because agents stop repeating the same paragraph-level advice without a project-level register. The skill deliberately avoids logging every minor pacing slip—single-instance problems stay in critique output, while patterns get names, severity, and cross-reference so your next drafting sprint has a checklist. It pairs naturally with critique and analysis skills in creative-writing-skills repos and stays useful from first manuscript discovery through post-launch content maintenance when you add chapters or spin-offs.1installs17Writing Principleswriting-principles packages evidence-based creative-writing guidance from the haowjy/creative-writing-skills repo, anchored in peer-reviewed work on transportation, reading flow, and social cognition rather than generic style tips. Solo builders shipping landing narratives, email sequences, founder stories, or interactive fiction can use it when copy must hold attention and feel human. The skill emphasizes measurable constructs from consumer and psychology research and flags where replication is contested, which helps indie teams avoid myth-driven writing advice. It pairs well with content planning in Validate and distribution work in Launch while living primarily on the Grow content shelf.1installs18Writing StaffingWriting-staffing is a creative-writing orchestration skill that explains which specialized agents should own which artifacts in a fiction or long-form pipeline. Instead of asking one general assistant to improvise an entire novel, you treat prose as production: the outliner structures arcs, chapters, and beats (often with mermaid maps), then a dedicated writer turns scene briefs into drafts grounded in style files and continuity anchors. The skill stresses operational discipline—parallel writers on the same scene degrade voice, so staffing is one writer per scene or chapter, with output always staged as draft material for critique and revision rather than publish-ready copy. Writers can run autonomously into a drafts directory or interactively when you want to co-write passages. For solo builders shipping newsletters, game lore, or indie fiction as part of their brand, this pattern mirrors how software teams separate planning from implementation: structure first, execution second, quality gates last.1installs