learn-recipe

/learn-recipe

Analyze recipe JSON and append every finding directly into the organization knowledge layer org/docs/. We don't accumulate findings in an intermediate file — we grow each doc directly.

All writes target org/docs/<relative-path>; the kit's docs/ is left alone. See @.claude/rules/org-knowledge-overlay.md.

Usage

• /learn-recipe <file-path> — analyze a specific recipe
• /learn-recipe <project-name> — analyze every recipe in a project
• /learn-recipe — analyze every recipe across every project

What we learn and where it goes

Every write target is org/docs/<relative-path>. Read the kit's docs/<same-relative-path> first and skip anything already documented there (only write differences, corrections, and org-specific additions).

1. Field info (most important)

Extract input/output field schemas from each step.

Sources:

• extended_output_schema — action / trigger output fields
• extended_input_schema — action / trigger input fields
• input — the actual input value (including datapill references)
• parameters_schema_json — Genie / Function parameters (JSON string)
• result_schema_json — Genie / Function result (JSON string)
• input.input.schema — Custom Action request schema
• input.output — Custom Action response schema

Destination: the matching action / trigger section of org/docs/connectors/<provider>.md.

Format:

### <action_name>

#### Input fields
| Field | Type | Required | Description |
|---|---|---|---|
| field_name | type | Yes/- | label |

#### Output fields
| Field | Type | Description |
|---|---|---|
| field_name | type | label |
| parent.child | type | label (nested) |

2. New providers / actions

When you find an unknown provider/name combination in a recipe (i.e. an action that doesn't appear in the kit's docs/connectors/<provider>.md).

Destination:

• Pre-built connectors → the trigger / action list table in org/docs/connectors/<provider>.md.
• Workato-internal providers → org/docs/platform/workflow-apps.md or org/docs/platform/agent-studio.md.