Workato Recipe Analyzer

Preprocesses Workato recipe JSON exports (1K–97K lines) into focused view files, stripping ~90–98% of UI/schema bloat to expose the actual recipe logic.

Workflow

1. Resolve the recipe path

Determine the .recipe.json path from:

• The argument (e.g., /workato-recipe global-context/sources/workato/.../str_rec_001...recipe.json)
• A path mentioned in conversation context
• If ambiguous, ask the user or use Glob to search: global-context/sources/workato/**/*.recipe.json

2. Run preprocessing

uv run python "${CLAUDE_SKILL_DIR}/scripts/extract_views.py" "$RECIPE_PATH"

The script outputs the views directory path to stdout. Views are cached in .scratch/workato-views/ (invalidated by source mtime or extractor version change).

Use --force to regenerate. Use --all to process all non-archived recipes.

Developer utilities live at the plugin root cli.py entry point. Consumers should call that stable CLI rather than depending on internal script paths.

3. Select analysis depth

Based on the argument or question:

overview (default — no depth specified):

• Read summary.json from the views directory
• Report: recipe name, version, trigger, connections, block/provider counts
• For callable recipes, show parameters and results

deep (argument contains "deep", a specific question, or a block reference):

• Spawn the workato-recipe:recipe-analyzer subagent with the views directory path and question
• The subagent reads summary.json and unified_logic.md to provide detailed analysis

View Files

File	Content	Size (typical)
summary.json	Metadata plus structured control-flow and error-handling contract	~4-8KB
unified_logic.md	Single unified document with control flow, mappings, variables, and error handling	~20-30KB

References

• ${CLAUDE_SKILL_DIR}/references/recipe-schema.md — Recipe JSON structure and datapill syntax
• ${CLAUDE_SKILL_DIR}/references/view-formats.md — View file formats and cross-referencing