Build an Interactive Story

Create an interactive educational story using $ARGUMENTS.

Follow the full conventions in stories/CLAUDE.md. The story must be self-contained HTML with inline CSS/JS, 10-15 scenes, 3-5 branching points, 2-3 endings, and keyboard navigation (1-9 keys).

Phase 0 — Plan (interactive)

1. Parse the argument: first word is the style name, optional remaining words hint at the topic.
2. Read the Style Guide Catalog table in README.md to confirm the style is available (empty Story column) and see what's already taken.
3. Read the chosen style's HTML file from styles/{name}.html to extract the design system.
4. Read stories/CLAUDE.md for the template and conventions.
5. Read 1 existing story to match structure — skim for the STORY data format, engine pattern, and component mapping.
6. If no topic was given:
   • Research the style's cultural/historical period
   • Propose 2-3 story concepts that thematically match the aesthetic
   • Ask the user to pick one
7. Propose the story arc: title, setting, 10-15 scene outline with branching diagram, key historical facts, and how style components map to narrative functions (e.g., "cards = case files", "alerts = discoveries").
8. Discuss with the user before proceeding.

Phase 1 — Research (parallel Sonnet subagents)

Launch 2-3 sonnet subagents in parallel to research:

• Historical facts: Real names, dates, numbers, events for the chosen topic. Need 8-12 concrete facts.
• Cultural context: Period details, atmosphere, vocabulary appropriate to the era.
• Style component mapping: How to repurpose the style guide's CSS classes for narrative elements.

Each subagent returns structured research notes.

Phase 2 — Build (sequential Opus subagents)

Use 3-4 sequential opus subagents:

1. First agent: Create the full HTML file with adapted CSS, title screen with back-to-stories link, scene wrapper structure, and the story engine JavaScript. Write the first 5 scenes of the STORY object. Include CSS for scene types (narration, investigation, confrontation, ending).
2. Second agent: Read the file, add scenes 6-10 (or remaining scenes). Ensure branching points connect correctly.
3. Third agent: Read the file, add remaining scenes and endings. Verify all next values point to defined scenes. Ensure every scene has a fact property. Test that no orphan scenes exist.
4. Fourth agent (polish): Read complete file. Verify scene graph integrity (no dead ends, no orphans), keyboard navigation works, clue tracker collects facts, all historical content is accurate.

Critical rules for build agents:

• Each agent MUST read the current file before writing
• All next values in choices must reference existing scene IDs in the STORY object
• Every scene needs: year, era, content, fact, choices
• Endings should offer "Play Again" (next: 'intro') and "Back to Stories" (next: '_exit')
• Back-to-stories link on title screen: <a href="index.html">

Phase 3 — Brief & Index

1. Write a research brief to stories/briefs/{style-name}-brief.md documenting the historical facts, story arc, and component mapping.
2. Read stories/index.html to understand the card format.
3. Add a card for the new story with: scene count, ending count, style name, title, description.
4. Update the Style Guide Catalog table in README.md — add the story title in the Story column for the chosen style.
5. Commit with message: Add {title} interactive story ({style-name} style)
6. Push using: git config --global credential.helper store && echo "https://GGPrompts:$(gh auth token --user GGPrompts)@github.com" > ~/.git-credentials && git push origin main