# Narrative Repo Setup > Set up a repository for fictional writing skills — choose the repository anchors, create the facet directories and domain documents, preserve existing project files, and report the files created. - Skill: `zenobi-us/narrative-repo-setup` (Agent Skill) - Install (CLI): `npx skillmds@latest add zenobi-us/narrative-repo-setup` - Raw SKILL.md: https://api.skillmd.com/api/skills/zenobi-us/narrative-repo-setup/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: zenobi-us (https://skillmd.com/u/zenobi-us) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/zenobi-us/narrative-repo-setup --- # Narrative Repository Setup Set up the repository before the other fictional-writing skills use it. Create a small, composable Markdown model for stories, characters, places, factions, technology, history, terms, and timeline events. Do not build a database. Do not copy copyrighted source text. Do not replace existing project instructions or story files. ## When to Use Run this skill once when a repository does not contain the narrative anchors. Run it again only to repair missing anchors or add missing facet directories. Do not use this skill to write a story, design a faction, or resolve a canon conflict. ## Repository Anchors Use these files as the single sources of truth: | Anchor | Purpose | |---|---| | `CONTEXT.md` | Project premise, scope, tone, core truths, canon boundary, and domain terms. | | `docs/adr/` | Decisions that are hard to reverse or surprising without their reason. | | `docs/agents/narrative-domain.md` | Facet types, metadata, IDs, links, and status values. | | `docs/agents/narrative-workflow.md` | Storage paths, update rules, and the writing lifecycle. | | `AGENTS.md` or `CLAUDE.md` | Pointer to the fictional-writing skills and their anchor documents. | If the repository has an active `ALIGNMENT-ROOT.md` or shared-agent context, resolve `CONTEXT.md`, `docs/adr/`, and `docs/agents/` against that alignment root. Keep facet and story files in the repository root. ## Facet Model Store one Markdown file per reusable facet. Use these directories: ```text characters/ stories/ timeline/ world/ places/ factions/ technology/ history/ terms/ ``` Use the existing directory layout when it already contains equivalent paths. Do not move files during setup. Every facet file starts with frontmatter like this: ```yaml --- id: faction.example type: faction name: Example Faction status: active canon: original era: red --- ``` Use stable IDs in the form `.`. Link related facets with normal relative Markdown links. Use `status: draft`, `active`, `retired`, or `archived`. Valid types are: - `story` - `character` - `place` - `faction` - `technology` - `history` - `term` - `event` ## Process ### 1. Resolve the roots Find the Git repository root. Read the root `AGENTS.md`, `CLAUDE.md`, and `README.md` when they exist. Read `ALIGNMENT-ROOT.md` when it exists. If shared-agent context defines an alignment root, use that root for alignment documents. Read `CONTEXT.md` and relevant files in `docs/adr/` before you edit them. Preserve existing terms and decisions. ### 2. Inspect before writing Check which anchors and directories already exist. Treat an existing file as authoritative. Do not overwrite it. Do not merge content automatically. If an existing file uses a different but clear location for an anchor, document that location in `docs/agents/narrative-workflow.md` instead of moving the file. ### 3. Create missing directories Create only missing directories: ```text characters/ stories/ timeline/ world/places/ world/factions/ world/technology/ world/history/ world/terms/ docs/adr/ docs/agents/ ``` Do not create example stories, characters, factions, or other fictional facts. ### 4. Create `CONTEXT.md` when absent When `CONTEXT.md` does not exist, create it with this content: ```markdown # Narrative Context This repository stores original fictional writing and its shared setting data. ## Premise _Record the central situation in one or two sentences._ ## Tone - _Record the intended tone._ ## Time and place - **Era:** _Record the era._ - **Primary location:** _Record the location._ - **Scope:** _Record the scope._ ## Core truths - _Record facts that every story must respect._ ## Canon boundary - **Official canon:** _Record which licensed setting material applies._ - **Original canon:** _Record that this repository's additions are original work._ - **Conflict rule:** _Record which source takes priority when sources conflict._ ## Open questions - _Record unresolved choices._ ## Domain terms - _Add one approved project term per entry._ ``` When `CONTEXT.md` exists, leave it unchanged. Report that it needs review if it lacks a canon boundary or domain terms. ### 5. Create the domain document when absent Create `docs/agents/narrative-domain.md` with the facet model, metadata, IDs, links, valid types, and status values from this skill. State that Markdown facet files are the source of truth. State that `CONTEXT.md` owns project language and `docs/adr/` owns hard-to-reverse decisions. ### 6. Create the workflow document when absent Create `docs/agents/narrative-workflow.md` with these rules: ```markdown # Narrative Workflow ## Source of truth - `CONTEXT.md` stores project scope, core truths, canon boundaries, and domain terms. - `docs/adr/` stores hard-to-reverse decisions. - Facet files store reusable setting and story data. - `timeline/` stores dated events and their sources. - Story files store story-specific arc and continuity notes. ## Storage - Store characters in `characters/`. - Store stories in `stories/`. - Store events in `timeline/`. - Store places, factions, technology, history, and terms in their directories under `world/`. ## Update rule Create or update a facet when a story introduces a fact that another story can reuse. Link the facet to the story that introduced it. Record a dated consequence in `timeline/` when a story changes the shared world. Record an open question instead of inventing an answer when the source material or project context does not decide the fact. Run continuity review after a story changes a character, faction, place, technology, or timeline event. ``` ### 7. Add the skill pointer Add a short `## Fictional writing` section to the existing `AGENTS.md` or `CLAUDE.md`. Do not create the other file when one exists. Use this text: ```markdown ## Fictional writing This repository uses the fictional-writing skills. Read `docs/agents/narrative-domain.md` for facet types and `docs/agents/narrative-workflow.md` for storage and update rules. Read `CONTEXT.md` before changing shared setting facts. ``` If the file already contains a fictional-writing section, update only stale paths. Do not add a duplicate section. ### 8. Report the result Report: - The repository root. - The alignment root, when one exists. - Existing anchors that were preserved. - New anchors and directories. - Existing anchors that need human review. - Any unresolved path or canon question. ## Common Mistakes | Mistake | Correct action | |---|---| | Create a second `CONTEXT.md` | Use the active alignment root and report the path. | | Overwrite an existing `CONTEXT.md` | Preserve it and report missing sections. | | Put facet data in one large YAML file | Use one Markdown file per reusable facet. | | Create sample fictional facts during setup | Create directories and templates only. | | Move existing stories or world notes | Keep existing files in place. Record their paths. | | Put process rules in `CONTEXT.md` | Put storage rules in `docs/agents/narrative-workflow.md`. | | Create an ADR for the setup itself | Create an ADR only when the repository makes a hard-to-reverse decision. | | Copy official setting text | Record a source and write a short summary. | | Create duplicate skill pointers | Update the existing section. | ## Completion Criteria The setup is complete only when: - The active repository and alignment roots are identified. - Existing instruction, context, and ADR files are preserved. - The required facet directories exist. - `CONTEXT.md` exists or its existing path is reported. - `docs/agents/narrative-domain.md` exists. - `docs/agents/narrative-workflow.md` exists. - The active agent instruction file points to the two narrative documents. - No fictional facts were invented. - No existing project file was overwritten or moved. - The report lists every file created, preserved, or requiring review.