slideflow-yaml-authoring

name	slideflow-yaml-authoring
description	Author and validate production-safe Slideflow YAML for Google Slides, Google Docs, Google Sheets, and local PowerPoint, including composable dbt connectors and AI providers.
version	2.0.0
spec	open-skill-v1

Slideflow YAML Authoring Skill

Purpose

Use this skill to produce production-safe Slideflow YAML and deterministic run commands across artifact kinds (slides, docs, sheets, powerpoint).

Supported provider ids:

• google_slides
• google_docs
• powerpoint
• google_sheets

Use this skill when

• creating or editing Slideflow config for Google Slides, Docs, Sheets, or PowerPoint
• translating reporting intent into data sources, transforms, replacements, and charts
• selecting between template, plotly_go, and custom charts
• defining AI summaries/replacements across providers (OpenAI, Gemini, Databricks serving)
• validating provider contracts before runtime failures

Required inputs

• target artifact kind (slides, docs, sheets, or powerpoint)
• provider configuration (template id/folder ids/credentials behavior)
• data source details (dbt/warehouse or direct sql/csv/json)
• runtime parameter shape ({param} tokens and optional params CSV)
• preferred AI provider (if AI summaries are required)

Deterministic workflow

1. Build the root contract for the artifact kind.
2. Define data sources with explicit names.
3. Prefer type: dbt + dbt: + warehouse: for new work.
4. Add replacements/charts (or workbook tabs for sheets) with explicit function names.
5. Validate references:
   • $column references map to real output columns.
   • every *_fn exists in function_registry.
   • template ids/placeholders/markers are valid for provider.
6. Run commands from references/provider-command-matrix.md.
7. Run gotchas pass from references/gotchas.md.

Guardrails

• Prefer additive, backward-compatible changes.
• Keep legacy databricks_dbt examples clearly marked as legacy-only.
• Keep secrets in env vars; never hardcode credentials/token values.
• Use explicit template ids and registry paths when ambiguous.
• For Databricks serving AI provider, use serving-endpoints base URL (not /invocations).
• For Google Sheets, define tab mode intentionally (replace, append).
• For citations-enabled slides configs, keep source metadata stable and deterministic.

Output contract

When producing config output, include:

• complete YAML for the selected artifact kind
• assumptions list (provider, runtime params, env vars)
• exact doctor/validate/build commands
• env var contract for data + AI provider
• note on whether templates are built-in or local

References

• references/config-schema-cheatsheet.md
• references/provider-command-matrix.md
• references/ai-provider-cheatsheet.md
• references/citations.md
• references/sheets-modes.md
• references/template-authoring-contract.md
• references/plotly-parameter-lookup.md
• references/gotchas.md
• assets/snippets/connectors.yml
• assets/snippets/replacements.yml
• assets/snippets/charts.yml
• .github/skills/slideflow-yaml-authoring/assets/examples/