playbook-generation

Playbook Generation Skill

TL;DR

• Generates a self-contained, branded HTML playbook from a topic and optional source files.
• Covers 13 AI-native workflows, 4 hands-on katas, progressive adoption timeline, and a value traffic light.
• Outputs a single HTML file (80-200KB) that works offline, prints cleanly, and looks professional.
• Follows the Sofka DS v5.1 design system with 42 CSS tokens and 22 components.
• Promotes the 3-layer grounding architecture (Drive/NLM/Gemini or adapted stack).

Inputs

• Topic (required): The subject of the playbook (e.g., "AI-native Project Management", "Automatizacion de QA con IA").
• Source files (optional): Directory or files containing context about the team, tools, problems, and processes. Supported formats: .md, .txt, .json, .yaml, .csv.
• Intake answers (gathered interactively): Audience/roles, tools/platforms, top problems, centralization strategy, language.

Outputs

• Primary: outputs/playbook-<topic-slug>-<YYYYMMDD>.html — Self-contained HTML playbook.
• Intermediate: outputs/.playbook-context.json — Extracted context from source files.
• Intermediate: outputs/.playbook-manifest.json — Structured content manifest (JSON) before HTML assembly.

Pipeline Overview

1. Ingest: Scan source files for roles, tools, problems, processes, metrics, and org info. Write to .playbook-context.json.
2. Clarify: Run intake questions to fill gaps not covered by source files. Build a structured brief.
3. Generate Manifest: Using the brief + context, produce a complete JSON manifest with all section content following prompts/content-generation.md.
4. Assemble HTML: Transform the manifest into a single-file HTML playbook with inline CSS, responsive layout, and print styles.
5. Validate: Check for unreplaced placeholders, size within 80-200KB, all required sections present, valid structure, consistent terminology.
6. Deliver: Report output path, file metrics, and offer browser preview.

Design System: Sofka DS v5.1

• 42 CSS custom properties (colors, spacing, typography, shadows, radii).
• 22 reusable components: hero, section-card, tool-card, kata-card, flow-card, timeline-step, decision-row, traffic-light, checklist, metric-badge, quote-block, callout, nav-bar, footer, print-header, and more.
• Responsive: Mobile-first with breakpoints at 768px and 1024px.
• Print-ready: @media print rules for clean A4 output.
• Accessibility: Semantic HTML, ARIA labels, sufficient contrast ratios.

Architecture: 3-Layer Grounding

Every playbook promotes a centralized information architecture to prevent AI hallucination:

• Capa de Datos: Google Drive (or Confluence/SharePoint/Notion) with governed naming and structure.
• Middleware: NotebookLM (or custom RAG) as anti-hallucination layer — only answers from your sources.
• Front: Gemini + Gems (or ChatGPT/Claude) as exploration and command center.

The architecture adapts to the user's actual stack while preserving the grounding principle.

Quality Gates

• No {{placeholder}} tokens remain in final HTML.
• File size between 80KB and 200KB.
• All 13 required sections present in the manifest and HTML.
• Consistent use of tool names (no mixing "Google Drive" and "GDrive").
• Valid HTML5 structure.
• All internal links resolve.

References

Reference files in the references/ directory provide templates, schemas, and examples:

• references/content-manifest-schema.json — JSON schema for the manifest
• references/brand-tokens-sofka.json — Brand design tokens (Sofka DS v5.1)
• skills/playbook-generation/references/section-templates.md — HTML templates per section type