Agent Skills for planning, tracking, and drafting fiction in markdown.

Story Skills gives agents a shared project format for fiction: a story bible, characters, worldbuilding, factions, artifacts, plot arcs, scenes, continuity state, promises and payoffs, timelines, and chapter drafts. Everything is plain markdown with YAML frontmatter, packaged as standard Agent Skills with Codex and Claude Code plugins.

The companion story CLI treats the story bible as a checkable contract. Its continuity engine catches dead characters walking, payoffs that land before their setup, unfired Chekhov guns, and stale story state, deterministically, before a reader finds them.

Why it works this way: Story Skills: a continuity compiler for AI-written fiction

Quick start

Install the plugin in Codex or Claude Code:

# Codex
codex plugin marketplace add danjdewhurst/story-skills
codex plugin add story-skills@story-skills

# Claude Code (type these inside a Claude Code session, not a shell)
/plugin marketplace add danjdewhurst/story-skills
/plugin install story-skills@story-skills

For any other agent that supports SKILL.md, use the Agent Skills CLI:

npx skills add danjdewhurst/story-skills   # or: bunx skills add danjdewhurst/story-skills

Then ask your agent to "Start a new story". Per-agent instructions for GitHub Copilot, Cursor, Windsurf, Gemini CLI, OpenCode, and others are under More install options.

New to Story Skills? Getting started walks through a first session, and the documentation index links every guide and reference page.

Or let your agent install it

Paste this prompt into your coding agent. It works out which agent it is and uses the matching install method:

Install the Story Skills bundle from https://github.com/danjdewhurst/story-skills.

First, work out which agent you are, then use the matching method below. If a command fails or you can't run it, tell me the exact command to run myself.

- Claude Code: run `claude plugin marketplace add danjdewhurst/story-skills`, then `claude plugin install story-skills@story-skills`. If the `claude` CLI isn't available, tell me to type `/plugin marketplace add danjdewhurst/story-skills` and then `/plugin install story-skills@story-skills` in this session.
- Codex: run `codex plugin marketplace add danjdewhurst/story-skills`, then `codex plugin add story-skills@story-skills`.
- Gemini CLI: run `gemini skills install https://github.com/danjdewhurst/story-skills.git`.
- Any other agent that supports SKILL.md (GitHub Copilot, Cursor, Windsurf, OpenCode, and others): run `npx skills add danjdewhurst/story-skills`, or `bunx skills add danjdewhurst/story-skills` if only Bun is installed. If that doesn't support you, clone the repository to a temporary directory and copy every folder in its `skills/` directory into your skills directory:
  - GitHub Copilot: `.github/skills/` in this project, or `~/.copilot/skills/` globally
  - Cursor: `.agents/skills/` in this project
  - Windsurf: `.windsurf/skills/` in this project, or `~/.codeium/windsurf/skills/` globally
  - OpenCode: `.opencode/skills/` in this project, or `~/.config/opencode/skills/` globally
  - Anything else: your documented skills directory, or `.agents/skills/` in this project

Prefer a project install unless I asked for a global one. If you can't tell which agent you are, ask me before installing. When you're done, tell me what you installed, where it went, and whether I need to restart or reload you for the skills to show up.

The continuity engine

Long-range consistency is what language models are worst at, and prompting can't fix it. Story Skills makes it deterministic: character deaths, promises and payoffs, open questions, scene casts, and durable knowledge and object state live in frontmatter, and story continuity treats contradictions the way a compiler treats type errors.

examples/the-unraveled-thread/ is a deliberately broken mystery. Every file is well-formed, so it passes story validate and story links cleanly, but the story itself doesn't hold together:

$ story continuity examples/the-unraveled-thread
Continuity check failed: 4 errors, 3 warnings, 0 dismissed
error: chapters/chapter-04.md lists edran-vale, who died in chapter-02; move posthumous appearances to mentions
error: continuity/promises/the-broken-compass.md pays off in chapter-02 before it is planted in chapter-03
error: continuity/questions/who-burned-the-mill.md resolves in chapter-02 before it is introduced in chapter-03
error: continuity/state.md knowledge-state[0] references missing chapter chapter-05
warning: chapters/chapter-03.md POV character nessa-thorn is not listed in characters [pov-not-in-cast]
warning: continuity/promises/the-sealed-letter.md was planted in chapter-01, 3 chapters ago, and has no payoff yet [promise-unpaid]
warning: continuity/state.md object-state[0] status active conflicts with worldbuilding/artifacts/vales-compass.md status destroyed [state-status-conflict]

Every finding is exact, file-addressed, and reproducible, and CI asserts this output on every commit. Intentional flashbacks and posthumous appearances stay legal through the chapter mentions field, and findings listed in continuity/exemptions.md are reported as dismissed. story doctor and story next fold the same checks into prioritized repair actions.

Skills

| Skill | What it does | Try saying | |-------|-------------|------------| | premise-workshop | Turns a spark into a tested premise: what-ifs, logline tests, premise and counter-premise, stakes, choice of form, titles, and comparable books, then hands off to story init --form | "Is there a novel in this idea?" | | story-init | Scaffolds the story bible, folders, and registries | "Start a new story" | | character-management | Creates character profiles with relationships, traits, arcs, and family trees | "Create a character" | | worldbuilding | Builds locations and systems: magic, politics, technology, religion, and more | "Design a magic system" | | plot-structure | Plans arcs with structures like three-act, hero's journey, Save the Cat, and kishotenketsu | "Create a plot arc" | | theme-craft | Builds the controlling idea (value + cause premise), the moral argument, lie/truth arc types, antagonist design, and motif/symbolism audits | "What's my story really about?" | | genre-craft | Genre packs with checkable conventions: mystery fair-play, romance beats, thriller, horror, MG/YA, sci-fi, and serial/episodic structure | "Plan a fair-play mystery" | | research | Investigates the real-world facts a story relies on, with quoted sources, confidence, interviews and site visits, invented-versus-accurate decisions, and qualified reviewers for legal, medical, and other risky details | "Fact-check the sailing in chapter 4" | | chapter-writing | Drafts chapters through an outline-first workflow that pulls from story context | "Write the next chapter" | | discovery-drafting | Pantsing mode: draft from a story kernel, keep post-hoc chapter notes, and reconcile the bible after each discovery-drafted chapter | "I want to discovery-write" | | scene-craft | Plans and checks the scene unit: Scene/Sequel structure, try/fail cycles, scene cards, dialogue subtext and voice differentiation, deep POV, exposition, flashbacks, and openings | "Does this chapter breathe?" | | voice-style | Keeps a copyeditor's style sheet (dialect, house spellings, dialogue punctuation, character voices, watch words) and acts on story prose lint findings | "Set up a style sheet for this book" | | verse-craft | Writes, scans, and fixes verse: limericks, sonnets, haiku, ballads, song lyrics, rhyming picture-book text, and a character's song or prophecy, with every line's stresses and the rhyme scheme shown so the author can check them | "Does this limerick scan?" | | line-editing | Owns the prose pass without flattening the author's voice: line edits with a reason for each change, distinct character voices checked by story voices, a copyedit against the style sheet, and read-aloud and proof passes | "Line edit chapter 3. Everyone sounds the same." | | revision-continuity | Revises drafts, audits continuity, and keeps character state, timeline, and arc changes consistent | "Continuity-check chapter 3" | | feedback-triage | Collects alpha/beta reader feedback per round, synthesizes convergent and divergent notes, and hands a revision plan to revision-continuity | "Triage the beta feedback" | | editorial-review | Handles work with other people: sensitivity and authenticity reader briefs, a real-people and defamation check, permissions for lyrics and epigraphs, the AI-use statement, rounds with human editors, review copies, and co-authoring | "I'm quoting a song lyric as my epigraph. What do I need?" | | series-continuity | Starts sequels and prequels as linked projects, carries characters and world forward, and checks shared canon across books | "Start a prequel to The Last Ember" | | submission | Checks submission readiness, drafts the query letter, pitch, comp titles, synopsis, and blurb, builds the Shunn manuscript, and tracks queries and responses | "Help me query agents" | | publishing | Self-publishing production: retailer metadata, ISBNs, the copyright page, accessible EPUB and print interiors, distribution, pricing, a launch plan with ARCs and ad testing, and a rights inventory with contract red flags | "Get my book ready for KDP and IngramSpark in 6x9" | | adaptation | Carries the story into other forms: an audiobook narration script with a pronunciation guide, a Fountain screenplay, a picture-book spread plan with illustration briefs, a comics script, an interactive-fiction branch map, and translated editions | "Make a narration script so I can audition narrators" | | story-maintenance | Runs deterministic CLI checks for validation, continuity, reports, indexing, links, word counts, import, and export | "Validate my story project" |

line-editing owns the prose pass. For general-purpose writing checks as well, add better-writing. It adds voice calibration, anti-generic writing checks, and a final prose-quality pass, and installs the same way:

npx skills add forjd/better-writing

Companion CLI

The optional story CLI handles deterministic project maintenance while the skills handle the creative work. It needs Node 18 or newer and has no runtime dependencies. Run it with npx, or install it globally:

npx story-skills --help
npm install -g story-skills   # then: story --help

To try unreleased changes, run it straight from GitHub with npx --yes --package github:danjdewhurst/story-skills story --help.

From a clone, use bun install and then bun run story -- --help. Copied-skill installs don't need either: story-maintenance bundles a scripts/story.js fallback that agents run with Node.

The CLI is for maintenance only. Agents write story content directly to markdown files and never create project-local build or generator scripts to emit the story.

Create and restructure

| Command | Purpose | |---------|---------| | story init "The Last Ember" | Scaffold a story project with the standard markdown layout | | story init "Book Two" --follows the-last-ember | Scaffold a sequel (or a prequel with --precedes) linked to an existing book, writing the backlink | | story import draft.md --title "The Lost Coast" | Split an existing manuscript into a new story project and suggest entity candidates | | story add character "Sera Voss" | Create entity files for characters, locations, systems, factions, artifacts, arcs, chapters, scenes, questions, promises, clues, terms, research notes, and matter pages | | story add matter "Dedication" | Add a front (default) or --placement back matter page such as a dedication, epigraph, or acknowledgments | | story add character "Пётр" --id petr | Give the id by hand for a name the CLI cannot slug, so the name keeps its own script | | story names "Seren" "Kestrel Row" | Check candidate names against every name, alias, and glossary term before using them: clashes fail, look-alikes warn | | story rename character sera-voss "Sera Vale" | Rename an entity and update kebab-case references | | story move chapter chapter-03 --number 4 | Renumber a chapter, or move a scene with story move scene <id> --chapter <id>, renaming the files and rewriting every reference to the old id | | story remove promise old-setup | Remove an entity and scrub metadata references | | story migrate [path] | Upgrade a project to the current schema |

Check and repair

| Command | Purpose | |---------|---------| | story validate [path] | Check required files, schema version, YAML frontmatter, registries, and word-count warnings | | story links [path] | Check character, location, chapter, and arc cross-references and backlinks | | story continuity [path] | Check deterministic continuity contracts: deaths, promises and payoffs, questions, casts, durable state, and travel times along location routes | | story series [path] | Order linked sequels and prequels by chronology and check shared canon: deaths, casts, knowledge fact ids, names, and destroyed artifacts | | story reindex [path] | Rebuild registry tables from the current markdown files | | story wordcount [path] --write | Count chapter prose and update chapter frontmatter plus the chapter registry | | story doctor [path] | Show health checks with actionable repair steps | | story next [path] | Recommend the next deterministic writing or maintenance actions, including the next revision pass while revising | | story passes [path] --init | Record named revision passes (structure, character, theme, continuity, pacing, line, copyedit, proof) in story.md; --start and --done mark progress | | story report [path] --actionable | Summarize inventory and optionally include next actions |

Analyze

| Command | Purpose | |---------|---------| | story knowledge sera-voss --at chapter-03 | Show what a character knew at a chapter, from timeline-scoped knowledge state | | story context chapter-03 --budget 6000 | Pack the drafting context for a chapter or scene (style rules, POV knowledge, cast cards, open threads, recent scenes) into a token budget, with nothing from later chapters | | story timeline [path] | Show scenes in story-time order from their date/time (marking scenes told out of order), POV balance by words, and each character's presence and longest absence | | story prose [path] | Lint chapter prose: filter words, -ly adverbs, said-bookisms, echoes, sentence rhythm, repeated phrases, similar names, and style-sheet.md spellings and watch words | | story voices [path] | Fingerprint each character's attributed dialogue (sentence length, contractions, questions, signature words) and flag voice-avoid words and characters who sound alike. Only named speech tags and single-name action beats count; pronoun tags do not | | story pacing [path] | Tabulate scenes, sequels, scene outcomes (yes, no, yes-but, no-and), and chapter hooks; flag runs of easy wins, missing sequels, flat chapter endings, and length outliers | | story clues [path] | Draw the fair-play grid of clue plants and reveals by chapter; flag late plants, unplanted reveals, and red herrings never debunked | | story diagram relationships --path . | Print Mermaid source for the family tree and relationships, the location route map, the story-time timeline, the clue flow, or arcs by chapter | | story progress [path] --log | Report words against target-words, the deadline, and chapter targets; --log records the day's count in progress.md for pace and a projected finish | | story compare [path] --ref draft-1 | Compare chapters with an earlier draft (a git ref, or --against a copied project folder): word changes, added and removed chapters, and unchanged paragraphs; --anchor ch03-p12 finds where a review-copy paragraph is now |

Publish

| Command | Purpose | |---------|---------| | story synopsis [--pages 1\|3] [--out file] | Compress arcs into a mechanical 1- or 3-page synopsis | | story export [path] --out dist/manuscript.md | Combine front matter, chapters, and back matter into a single manuscript markdown file | | story build [path] --format epub | Build disposable markdown, EPUB, DOCX, or Shunn manuscript artifacts in dist/; EPUB builds embed the story.md cover image, publishing metadata, and accessibility metadata | | story build [path] --format html | Build a single-file review copy whose paragraphs carry citable anchors such as ch03-p12 | | story build [path] --format print --trim 6x9 | Build a print-ready paged-media HTML interior (author and chapter-title running heads, foot page numbers, raised chapter initials) to render to PDF with Paged.js, WeasyPrint, or Prince | | story build [path] --format narration | Build an audiobook narration script with a pronunciation guide and runtime estimates | | story build [path] --format metadata | Build a retailer metadata sheet from story.md with a readiness checklist | | story build [path] --format fountain | Build a Fountain screenplay skeleton, one scene heading per scene record, to draft an adaptation into |

Behavior notes:

• Matter pages from matter/ appear in the export and in every build format except Shunn, which is a submission format.
• EPUB and DOCX builds target plain prose: *italic* and **bold** become italic and bold runs, scene-break lines (***, ---) become a * * * separator, and other markdown structure such as blockquotes, lists, and tables is flattened to text. The markdown export keeps chapter text as-is.
• story rename, story move, and story remove update entity ids in frontmatter reference fields and markdown link targets. They never edit prose, so a character called "Port" can be renamed without touching the word "port" in chapter text.
• A command that changes a frontmatter value rewrites only the entries that changed. Comment lines, unchanged entries, and the body keep their exact text, and files whose values don't change are left untouched.

Every command and option is in the CLI reference. For a complete first session, read Getting started. For the project contract, read the Project format reference and schemas/story.schema.json.

Write a book with pull requests

A story project with deterministic checks is one an agent can advance unattended. The templates/github/ workflows turn a story repository into a self-drafting book:

• story-checks.yml runs story validate, story links, story continuity, and story report --actionable on every push and pull request, so a chapter PR can't merge with a continuity contradiction.
• draft-next-chapter.yml runs Claude Code on a schedule. It asks story next for the next action, drafts the next chapter with the chapter-writing skill, updates scene records and continuity state, runs the maintenance checks, and opens a pull request for review. When story next suggests no chapter (the story is revising or complete, or every arc is resolved), it stops without drafting.

Copy both files into .github/workflows/ in the repository that holds your story project, add an ANTHROPIC_API_KEY secret, and review one chapter PR each morning.

GitHub doesn't start story-checks.yml for pull requests opened with the built-in GITHUB_TOKEN, so the draft workflow runs the same checks itself after drafting. Pass a personal access token as github_token if you also want the checks workflow to run on those PRs.

Readers who never open a terminal can review the book too:

• review-copy.yml builds story build --format html on every push to main and publishes it to GitHub Pages. Every paragraph in the review copy carries a label such as ch03-p12, so a note can point at an exact place.
• ISSUE_TEMPLATE/manuscript-note.yml is an issue form that asks readers for that label, the kind of note, and how much it affected their reading. The feedback-triage skill turns those issues into a feedback round.

Copy the workflow into .github/workflows/ and the form into .github/ISSUE_TEMPLATE/, then set Settings > Pages > Source to GitHub Actions. Pages sites are public unless your plan supports private Pages; for a private manuscript, delete the deploy job and share the workflow artifact instead.

Import an existing manuscript

Most writers don't start from a blank page. story import builds a Story Skills project from work in progress:

story import draft.md --title "The Lost Coast" --genre mystery

It splits the manuscript on chapter headings (or imports a directory of chapter files in natural name order, so chapter-2 comes before chapter-10), creates the full project layout with accurate word counts and registries, and prints recurring proper-name candidates so an agent can follow up with story add character and story add location to build out the bible.

--force lets an import reuse an existing directory. It replaces every chapter-NN.md file in chapters/, so stale chapters from an earlier import are removed.

Project structure

Running story-init creates this layout:

my-story/
├── story.md                  # Story bible: title, genre, themes, POV, tense
├── style-sheet.md            # Voice, house spellings, and watch words
├── characters/
│   └── _index.md             # Character registry
├── worldbuilding/
│   ├── _index.md             # World overview
│   ├── locations/
│   ├── systems/
│   ├── factions/
│   └── artifacts/
├── plot/
│   ├── _index.md             # Arc overview
│   ├── arcs/
│   └── timeline.md
├── scenes/
│   └── _index.md             # Machine-readable scene registry
├── continuity/
│   ├── state.md              # Character, object, and knowledge state
│   ├── questions/
│   │   └── _index.md
│   ├── promises/
│   │   └── _index.md
│   └── clues/
│       └── _index.md
├── glossary/
│   ├── _index.md
│   └── terms/
└── chapters/
    └── _index.md             # Chapter registry

Some files appear only once you need them: matter/ for front and back matter, research/ for research notes, progress.md for the session log written by story progress --log, and continuity/exemptions.md for dismissed continuity findings.

How it works

Every story element is a markdown file with YAML frontmatter, and the skills cross-reference those files to keep the project consistent:

• story.md is the top-level bible that every skill reads. Its schema-version: 2 field lets the CLI detect incompatible project formats.
• Every entity file is named by a kebab-case identifier, such as sera-voss or chapter-01. It comes from the entity's name, or from story add --id when the name has no ASCII letters or digits; names themselves may be written in any script.
• _index.md files are the registries for each domain.
• Relationships and references are kept bidirectional.
• Scene records and continuity state keep character knowledge, object ownership, and setups and payoffs in files, so they carry over between sessions.

Examples

Complete projects generated with Story Skills:

• The Cormorant Tide
• Pippa and the Borrowed Star, a children's Christmas story (6 chapters, 2,183 words)

Examples in this repository:

• examples/the-last-ember/: a fantasy with three characters, two locations, a magic system, a plot arc with foreshadowing, and a drafted first chapter.
• examples/the-fall-of-the-citadel/: a prequel to The Last Ember, linked with series, book-number, and precedes, that shares characters and places with the first book. Run story series examples/the-last-ember to see the chronology.
• examples/harbor-of-second-light/: a near-future coastal mystery with memory technology, a posthumous witness arc, populated continuity state, and a drafted first chapter.
• examples/the-gull-rock-light/: a short branching story whose chapters carry choices; story build --format twee turns it into a Twine story and --format ink into an ink story.
• examples/the-unraveled-thread/: a deliberately broken project that demonstrates the main kinds of finding the continuity engine reports.

More install options

The plugin install in Quick start is the recommended path. For local skill authoring, copy the skills in directly; Codex detects repo and user skills automatically:

git clone https://github.com/danjdewhurst/story-skills.git

# User-wide
mkdir -p ~/.agents/skills && cp -r story-skills/skills/* ~/.agents/skills/

# Or repo-scoped
mkdir -p .agents/skills && cp -r story-skills/skills/* .agents/skills/

VS Code with Copilot discovers skills from several directories:

git clone https://github.com/danjdewhurst/story-skills.git

# Copy skills to your project (either works)
mkdir -p .github/skills && cp -r story-skills/skills/* .github/skills/
mkdir -p .agents/skills && cp -r story-skills/skills/* .agents/skills/

# Or install globally
mkdir -p ~/.copilot/skills && cp -r story-skills/skills/* ~/.copilot/skills/

Copilot can activate a skill when your request matches its description, or you can invoke one manually.

Cursor supports the SKILL.md standard:

git clone https://github.com/danjdewhurst/story-skills.git
mkdir -p .agents/skills && cp -r story-skills/skills/* .agents/skills/

Windsurf discovers skills from workspace and global directories:

git clone https://github.com/danjdewhurst/story-skills.git

# Copy skills to your project
mkdir -p .windsurf/skills && cp -r story-skills/skills/* .windsurf/skills/

# Or install globally
mkdir -p ~/.codeium/windsurf/skills && cp -r story-skills/skills/* ~/.codeium/windsurf/skills/

Cascade can invoke a matching skill automatically, or you can use @skill-name to invoke one directly.

Gemini CLI supports the same SKILL.md format through the Agent Skills standard:

# Install all skills globally
gemini skills install https://github.com/danjdewhurst/story-skills.git

# Or install a single skill (any folder under skills/)
gemini skills install https://github.com/danjdewhurst/story-skills.git --path skills/chapter-writing

# Or link locally after cloning
git clone https://github.com/danjdewhurst/story-skills.git
gemini skills link story-skills/skills

Gemini can activate a skill when your request matches its description.

OpenCode supports the SKILL.md format natively:

git clone https://github.com/danjdewhurst/story-skills.git

# Copy skills to your project
mkdir -p .opencode/skills && cp -r story-skills/skills/* .opencode/skills/

# Or install globally
mkdir -p ~/.config/opencode/skills && cp -r story-skills/skills/* ~/.config/opencode/skills/

OpenCode also searches common skill paths such as .claude/skills/, so it can find project-level skills installed for other agents.

These skills follow the open Agent Skills standard. If your agent supports the Agent Skills CLI, install the bundle directly:

npx skills add danjdewhurst/story-skills   # or: bunx skills add danjdewhurst/story-skills

Use --skill <name> to install only specific skills, or --agent <name> to target a supported agent. You can also copy the skill folders into any compatible agent's skills directory.

Outside coding agents:

• Claude.ai or ChatGPT Projects: add the SKILL.md and reference files as project knowledge.
• Any LLM API: include the skill content in the system prompt.
• By hand: the templates, workflows, and project structure are model-agnostic.

Development and releasing

To contribute, start with CONTRIBUTING.md and the Code of Conduct. User-visible changes are listed in the changelog. The Development guide covers the repository layout, CLI architecture, tests, and release process in full. Development uses Bun:

bun install
bun run test
bun run test:coverage
bun run test:examples   # also validates every example against schemas/story.schema.json
bun run check:metadata

bun run build:fallback generates the copied-skill fallback CLI from the package entrypoint. After changing CLI source, rebuild it, check it is current, and confirm it runs under Node:

bun run build:fallback
bun run check:fallback
node skills/story-maintenance/scripts/story.js --help

The evals/ harness regression-tests the writing skills. Fixtures seed a drafting brief with known canon and known traps. A dependency-free checker verifies that drafts keep the canon and spring none of the traps, a model runner (requires the claude CLI) drafts through a real model and judges for invented canon, and a pairwise comparison measures the skill against a no-skill baseline. See evals/README.md.

bun run check:evals      # validate fixture schemas
bun run eval:selftest    # checker self-test against known-good drafts
node evals/run-skill.js  # full model run (needs Claude Code credentials)

Every published change needs a new version in package.json, .codex-plugin/plugin.json (Codex's version source), .claude-plugin/plugin.json (Claude Code's), and src/version.js (printed by story --version), so installed users receive updates. Marketplace entries stay unversioned to avoid duplicate version state.

Don't bump these by hand. The release script bumps all four, plus the template STORY_REF pins and the version examples in the docs, moves the Unreleased entries in CHANGELOG.md under the new version (and refuses to run while there are none), rebuilds the fallback, runs the CI checks, commits chore: release X.Y.Z, tags vX.Y.Z, pushes, and creates a GitHub release with generated notes. The tag push runs the Publish workflow, which publishes the package to npm with provenance through trusted publishing. The script requires a clean main that matches origin/main, a logged-in gh, and a version that isn't already on npm:

bun run release patch            # or minor, major, or an explicit version like 1.2.0
bun run release patch --dry-run  # run the checks and print the plan without changing anything

Distribution metadata lives in .claude-plugin/ for Claude Code and in .codex-plugin/ plus .agents/plugins/marketplace.json for Codex. The plugins/story-skills symlink is intentional: Codex marketplace entries must point at a child plugin directory, so the symlink exposes the repo-root plugin without duplicating skills/.

License

MIT