Search skills/

Oh Story Claude skills

13 skills for writing Chinese web novels: ranking scans, teardown of hit books, long and short drafting with file-based continuity tracking, de-AI editing and covers.

zenstory-ai/oh-story-claudecode 13 skills Writing & editing
Our verdict Recommended with caveats

Deep, scripted and well-documented for its niche, but the browser-cdp skill copies Chrome login data and the ranking scrapers hit third-party sites, so enable those parts deliberately.

4/5 Depth expertise shipped ! Security Broad access 0d Maintenance Updated today 8.1K Adoption stars OK License MIT, commercial OK

What Oh Story does

Oh Story targets Chinese web fiction (Qidian, Fanqie, Jinjiang, Qimao, Zhihu Yanxuan). The README describes a pipeline: scan ranking lists to pick a theme, analyze successful books, build an outline and write chapters, remove AI-sounding phrasing, and generate a cover. It is split into 13 skills such as story-long-write, story-short-write, story-long-analyze, story-long-scan, story-deslop, story-review, story-import, story-cover and a /story router.

Its main idea is to use files as memory. Settings, outline, text and tracking live in separate files, and a JSON state file is the single source for derived views like the current-context card and list of open plot threads. Scripts such as storyctl.py check a chapter (length, AI patterns, degenerate output) and commit tracking state; hand-edited derived files are rejected by the check. /story-setup deploys hooks, agents and references for the host. Skills carry many reference files (the setup and write skills hold the most) and dozens of scripts, though no evals. The ranking scanners are Node scrapers for specific sites, and story-cover calls GPT-Image-2 (via Codex's built-in image tool or an API key). Hosts named: Claude Code, Codex, OpenCode, Antigravity, ZCode, OpenClaw and Reasonix.

Good for

  • Authors writing long Chinese web novels who need continuity across sessions
  • Studying hit books and ranking trends before outlining
  • Editing AI-sounding Chinese prose

Not for

  • Non-Chinese fiction writing
  • Anyone unwilling to run scripts that scrape sites or control Chrome

What's inside depth 4/5

13 skills averaging 12 KB with hundreds of reference and template files and dozens of scripts including chapter checks and state commit tools, but no evals.

Skills
13
Reference docs
359
Scripts
83
Tests / evals
0
Templates
30
Avg SKILL.md
12 KB

Counted from the repository file tree. How we score depth.

Dependencies and costs

  • Node.js 20+ and Python for the scripts
  • agent-browser (npm) and Chrome for browser-cdp and the ranking scanners
  • GPT_IMAGE_API_KEY for story-cover unless using Codex's built-in image tool
  • Windows users may need symlink support (core.symlinks) for Codex

Security review

Broad access

This skill asks for more access than most, or sends data to another service. That is usually a design choice rather than bad intent, but you should know before you install it.

  • Uses your browser cookies or logged-in profileCopies Chrome's Cookies, Login Data, Login Data For Account and Network/Cookies files into ~/chrome-debug-profile and launches Chrome with --remote-debugging-port (default 9222, localhost) on that profile so the agent can drive your logged-in sessions via agent-browser --cdp; SKILL.md also shows reading localStorage tokens and document.cookie. The ranking scanners (story-long-scan, story-short-scan) use this browser.skills/browser-cdp/scripts/setup-cdp-chrome.js:621

A Chinese web-novel writing suite with checkers, ranking scanners and a project setup skill. The browser-cdp skill copies your Chrome cookie and saved-login files into a second profile and exposes a localhost debug port so the agent can reuse your logged-in sessions; it asks before killing running Chrome. The setup skill installs hooks and a git pre-commit block into your project, and the cover skill can send an API key to an OpenAI-compatible endpoint.

Worth knowing

  • Makes network requestsScanner scripts load public ranking pages (Qidian, Fanqie, JJWXC, Qimao, Ciweimao, Heiyan and similar) through the CDP browser; story-cover sends prompts to an OpenAI-compatible images API; story checks api.github.com for a newer release.skills/story-long-scan/scripts/cdp-utils.js:1
  • Needs an API key for its own servicestory-cover's API fallback reads GPT_IMAGE_API_KEY and sends it as a Bearer header to the configured OpenAI-compatible base URL; Codex's built-in image tool is preferred when available.skills/story-cover/references/api-fallback.md:46
  • Installs with curl | shOpenCode upgrade instructions include curl -fsSL https://opencode.ai/v2/install | bash (also in references/deploy-opencode.md:11) as an alternative to npm i -g @opencode/cli.skills/story-setup/UPGRADING.md:52
  • Runs remote code at an unpinned versionInstall uses unpinned npx skills add for the skill pack, and browser-cdp requires npm install -g agent-browser.README.md:53
  • Installs packages or toolsNeeds agent-browser (npm global), Node.js 20+, Chrome, and curl/jq/base64 for the cover API fallback.skills/browser-cdp/SKILL.md:9
  • Relies on a CLI or package we could not reviewRelies on the agent-browser CLI, Chrome, git and the host agent CLIs (Claude Code, Codex, OpenCode, Antigravity); setup-cdp-chrome.js can also run pkill -9 / taskkill on Chrome after a documented consent step (--dry-run, then ask).skills/browser-cdp/SKILL.md:9

What we checked

The repository (808 files) on Oct 29, 2026. Not a full review: 428 of 808 files were not downloaded (large repository; we reviewed the skill folders); we read 5 of 71 code files; it relies on code outside the repository: agent-browser npm package, Google Chrome, and the host agent CLIs; Most of the 71 code files were not opened in full, including setup-cdp-chrome.js (41 KB), story_codex_hook.py, story_hook_core.js, the rank scrapers, tracking_commit.py, author_memory_commit.py, check-ai-patterns.js and the dashboard server; their network, spawn, copy and bind-address behavior was located by keyword search and partial reads only; About 428 readable files not in the local copy (.agents notes, plugin manifests such as .zcode-plugin, other docs).

NVIDIA SkillSpector static scan: 43 pattern hits (Dangerous Code Execution 12, Rogue Agent 8, MCP Rug Pull 7, Data Exfiltration 5, Data Flow 3, Tool Misuse 2). The 43 hits are mostly subprocess calls to git, node, fc-list and the checker scripts, 'write skill' / 'return prompt' phrases in Chinese-language workflow docs, 'rm -' temp-file cleanup in an API fallback snippet, unpinned npx mentions, and the opencode curl | bash line. The Chrome profile copying that matters is browser-cdp, which the scanner did not flag but the file inspection confirms.

A review, not a certification. Skills change after we check them — read the files before installing.

License

MIT — commercial use allowed.

Can it run unattended?

Not really. It has persistent state and chapter gates for resuming work, but the author decides direction and approves plans; it is designed for guided sessions rather than unattended runs.

Alternatives to Oh Story

Humanize KR (im-not-ai)

epoko77-ai/im-not-ai

A Korean-language humanizer that detects and rewrites AI-sounding text (translationese, mechanical parallelism) while preserving facts, with deterministic gates.

Recommended Check first Depth 4/5

Anti-Defensive Writing

Adkid-Zephyr/anti-defensive-writing-Skill

A lightweight CN/EN academic-writing skill that restructures paper drafts around one strongest claim instead of hedged, evenly-weighted self-critique.

With caveats No issues Depth 2/5

Drama Skills

zenstory-ai/drama-skills

Eleven skills that take an idea or novel through script, character assets, storyboard, image and video prompts and a rough cut for AI short dramas.

Recommended Check first Depth 5/5

Planning with Files

OthmanAdi/planning-with-files

Keeps task_plan.md, findings.md and progress.md on disk and re-injects them via hooks, so a plan survives /clear, compaction and crashes.

Recommended Broad access Depth 4/5

FAQ

What is Oh Story (oh-story-claudecode)?

It is a set of 13 agent skills for writing Chinese web novels, covering ranking scans, book teardown, long and short drafting, de-AI editing, review and cover generation. It keeps story state in files so work can resume after context is compressed.

How do I install Oh Story?

Run npx skills add zenstory-ai/oh-story-claudecode -y -g, then run /story-setup in your writing project (or $story-setup in Codex) and start a new session. The README includes host-specific install notes, including Windows ENOENT issues.

Is Oh Story safe to install?

Most of it is instructions and local file checks, but the browser-cdp skill copies Chrome cookie and login-data files into a debug profile so the agent can reuse logins. Skip or review that skill if you do not want an agent to reach your logged-in browser sessions.

Does it write in English?

No, it is built for Chinese web-novel platforms and conventions. Prompts, genre cards and the de-AI rules are Chinese-specific.

Sources: GitHub repository (stars, license, file tree, last push Oct 29, 2026). Review based on the README and SKILL.md files as published.