worldwonderer/oh-story-claudecode
13 skills33.4k installs27.2k starsGitHub
Install
npx skills add https://github.com/worldwonderer/oh-story-claudecodeSkills in this repo
1Story Long Writestory-long-write embeds the complete de-AI (去AI味) guide used after long-form chapter generation in the oh-story Claude Code suite. Solo fiction builders relying on agents for serial output hit platform and reader rejection when prose shows telltale LLM cadence, banned diction, triple-layer action stacking, or moralizing chapter endings. This skill is journey-wide process knowledge: route by task—self-check versus paragraph rewrite versus full-chapter three-pass cleanup—and apply fingerprint rules, quick-scan mnemonics, and conflict-dialogue rewrite exemplars. It explicitly defers vocabulary lists to banned-words.md while adding higher-order replacement principles for abstract fate metaphors and generic similes. Creative intent wins over mechanical synonym swaps when rules conflict. Use whenever long chapters need human voice consistency, not only at initial draft time.2.9kinstalls2Story Long Analyzestory-long-analyze is a long-form Chinese web-fiction teardown skill for agents. It treats blockbuster novels as research material: first it analyzes the golden three opening chapters and emits a quick preview report, then—if you confirm—it continues through per-chapter summaries, aggregated pacing and payoff patterns, setting relationships, and a consolidated report. All outputs are written to 拆文库/{书名}/, with mandatory original-text backup and resume via _progress.md when a run stops mid-way. Solo creators and indie fiction writers use it when they have a title, platform, or source file and want systematic craft insight instead of casual chapter notes. It is optimized for serial fiction workflows (爽点, 人设, 节奏) rather than SaaS code, but fits Prism’s Idea and Grow content lanes for builders who also ship newsletters, fiction IP, or scripted media.2.9kinstalls3Story DeslopStory Deslop is a checker-and-workflow skill for solo writers using Claude Code to produce Chinese long- or short-form fiction without the telltale cadence of model-generated prose. It packages a full anti-AI writing guide: high-frequency diction traps, forbidden summary endings, triple-stacked action paragraphs, and decision routing so you know whether to run a quick scan, a banned-word pass, or the systematic three-pass deslop method. The skill treats preservation of author intent as overriding mechanical synonym swaps—replacements must keep plot function and voice. It integrates with sibling story skills (long-write, short-write, review, setup) via a shared reference file guarded by CI sync scripts. Builders install it when drafts sound uniformly polished, overly philosophical at chapter ends, or emotionally told rather than shown. It does not replace plot structure or genre research; it is the editorial hygiene layer after generation.2.8kinstalls4Story Short WriteStory Short Write bundles a Chinese-language anti-AI writing playbook for fiction authors using Claude Code-style agents. It targets solo creators shipping short chapters or serial fiction who need prose that avoids obvious LLM fingerprints—summary endings, tell-not-show emotion, banned diction, and the same action stretched across redundant paragraphs. The guide routes you by intent: post-draft self-check, paragraph rewrite, banned-word scan, or full-chapter three-pass cleanup. It encodes seven detectable AI patterns, show-don’t-tell supplements, conflict-dialogue rewrite samples, and fast mnemonic checks aligned with sibling skills such as story-deslop and story-review. Install it when you finish a short-fiction segment and want a repeatable editorial pass before publish or platform upload. It complements setup and analysis skills in the same monorepo rather than replacing plot planning or long-form structure skills.2.8kinstalls5Story Short AnalyzeStory-short-analyze is a Chinese-language agent skill for fiction writers who use Claude Code to critique short-story prose for “AI flavor”: overused phrases, stacked redundant description, tell-not-show emotion, and formulaic chapter closings. It points the agent at a consolidated anti-AI writing reference with routing tables, high-risk word lists cross-linked to banned-words.md, rewrite exemplars, and a quick self-check mantra. Indie creators publishing serial fiction or brand stories install it right after a draft pass from story-short-write or similar skills, before human edit or release. It does not deploy infrastructure; it improves readability and authenticity of creative content. Use in Grow when refining pieces for audiences, and in Validate when testing whether AI-assisted drafts feel human enough to ship.2.8kinstalls6Story Long Scanstory-long-scan is an agent skill packaged as a Chinese web-fiction (网文) reference handbook for solo builders and indie authors who ship serial content with Claude Code-style agents. It concentrates on what to write before you write long: how to turn a金手指 or scene into a sellable core hook, how to choose题材 with roughly 85% familiarity and 15% surprise, and how to stay inside the three boundary circles so you do not break reader expectations. The skill documents practical 扫榜 playbooks for Qidian (三江, 首订图, 签约新书榜) and Fanqie (新书榜 with在读 thresholds), flags high-risk题材 when samples are thin, and summarizes流派 rules—极道加点, 凡人精打细算, 种田经营, 都市高武模板 chapters, and跟风换皮 strategies. Use it when you need structured market and craft context in chat instead of ad-hoc guessing; it is procedural knowledge, not a live crawler. Pair with your own notes and platform browsing when validating trends.2.8kinstalls7Story Short Scanstory-short-scan is a Chinese-language reference skill for cross-platform short web fiction. Solo builders and indie creators use it when they must choose among major channels such as 七猫, 黑岩, 知乎, and 点众, then write blurbs and pick genre stacks that match each site’s length, POV, dialogue density, and hook patterns. The SKILL routes you through a decision table so you compare platforms on dimensions like chapter count, paragraph rhythm, intro requirements, and爽点节奏 before you commit to a draft structure. It embeds platform-specific intro formulas, opening and closing motifs, and reversal tactics drawn from historical hot samples, with a hard editorial guardrail: treat those patterns as hypotheses until you reconcile them with a fresh market scan. For Prism’s audience shipping content products or agent-assisted fiction pipelines, it is editorial research and packaging logic—not a generic writing coach—optimized for marketplace-native serials rather than Western novel conventions.2.8kinstalls8Browser CdpBrowser CDP is a phase-specific agent skill built around a Node setup script that launches or prepares Google Chrome with the Chrome DevTools Protocol enabled, so tools like agent-browser can attach to a real browser instead of a sterile headless profile. Solo builders use it when automated flows need authenticated sites—dashboards, staging apps, or OAuth-heavy tools—without re-architecting login for every task. The script supports detect-only probes, dry runs, profile reset, and non-interactive confirmation flags, which makes it safer to invoke from CI-minded agents that cannot answer terminal prompts. It documents exit codes and structured stdout for readiness checks. Install Chrome locally, understand that restarting Chrome may be required, and treat killing existing browser processes as an explicit consent step the skill surfaces.2.8kinstalls9Story Coverstory-cover is a reference skill that encodes how major Chinese web-fiction platforms expect novel covers to look, and turns that into reusable English prompts for image models such as GPT-Image-2. Solo builders and indie authors who publish on Tomato, Qidian, Jinjiang, or similar channels can avoid generic AI art that fails platform norms. The readme defines visual language per platform—high-saturation mass-market layouts versus minimalist literary posters—and supplies keyword bundles agents can assemble into complete prompts. A keyword-to-genre table helps infer xianxia, urban, ancient romance, modern romance, and suspense aesthetics from title or blurb text. Use it when you need consistent, market-appropriate cover direction without hiring a designer for every variant, or when your agent must localize a Chinese title’s positioning into visual briefs for generative tools.2.5kinstalls10Story Setupstory-setup is the foundational reference in the oh-story-claudecode stack for Chinese fiction writers using Claude Code and similar agents. It centralizes how to detect AI writing fingerprints—stock diction, chapter-end summarizing, stacked redundant description—and routes you to the right module for self-check, targeted rewrite, or a full chapter pass. Solo creators serializing novels or interactive fiction often lose reader trust when drafts sound uniformly synthetic; this skill encodes editorial discipline (priority checks on tier-one banned words, context-sensitive replacements, and preserving story intent over mechanical swaps) so downstream story-deslop, story-review, and write skills share one consistent rulebook. Use it at project kickoff to align voice constraints, after any agent-generated chapter before publication, and whenever you need exemplar-level guidance for dialogue and sensory prose. It does not replace plotting skills; it makes generated prose credibly human on the page.2.1kinstalls11StoryStory is the main entry skill for the Oh-Story Claude Code toolbox aimed at Chinese web fiction authors. When a user says they want to write a novel, analyze rankings, remove AI tone, or import an existing manuscript—but does not name a concrete command—the router reads intent keywords, matches a dispatch table, and invokes the correct specialized skill such as long-write, short-write, long-scan, or story-deslop. It also enforces environment hygiene: missing 追踪/ or 设定/ trees trigger setup, and undeployed projects get story-setup first. For operators managing several titles, it lists book folders and handles switch/list flows. Solo creators treating fiction as a product get one conversational front door instead of memorizing a dozen slash commands, while power users still land directly on scan or analyze when keywords are explicit.2.1kinstalls12Story Reviewstory-review is a checker-oriented agent skill for solo builders and indie authors who ship serialized or long-form fiction with AI-assisted drafting. After chapters or scenes are written, the agent routes you through AI fingerprint detection—high-frequency words, abstract uplift, forbidden closings, and redundant action layering—and applies a systematic three-pass de-AI method with a banned-words quick reference. It is built for Claude Code workflows in the oh-story stack, where multiple writing skills share one maintained anti-AI guide. Use it when prose sounds generic, endings moralize, or dialogue reads flat; skip it for code, APIs, or marketing briefs that are not narrative fiction. The outcome is manuscript-ready prose that keeps plot function and author intent while reducing detectable LLM stylistic habits.2.1kinstalls13Story ImportStory-import is a specialized agent skill for solo novelists and serial fiction writers using Claude Code to manage long-form Chinese (and similar) projects. After拆书 analysis lands structured files under 拆文库, this skill reverse-engineers 追踪/角色状态.md entries from those artifacts only—deliberately not re-parsing full 原文—to honor analyze-then-migrate discipline. It enforces generation order so 伏笔.md exists before step six queries embedded foreshadowing, and 上下文.md comes last because it depends on recent character deltas. Each tracked character receives a seven-step reconstruction spanning current identity, abilities, relationships with evolution notes, public reputation from latest plot beats, and open foreshadowing tied to 已埋 entries. Short fiction is explicitly out of scope per state-tracking rules. The outcome is a consistent tracking layer your drafting agents can load every session, reducing continuity bugs across dozens of chapters. Confidence is moderate for English-only builders unless your repo already follows the oh-story folder conventions.1.9kinstalls