catalog

name	catalog
description	Scan the organization's shared assets (Recipe Functions, connections) and catalog them. Referenced by `$create-recipe` and `$plan`. Japanese prompts are also supported.

$catalog

Scans assets in the shared projects under the organization's projects/ and generates / updates a catalog file. Other skills ($create-recipe, $plan) consult the catalog to propose reusing existing assets.

Important: do not scan private-scope projects. Respect departmental visibility controls.

Usage

• $catalog — display the catalog
• $catalog scan — scan shared projects and regenerate the catalog
• $catalog config — view / edit scope settings
• $catalog add <file-path> — manually add an asset to the catalog

File layout

projects/
├── CATALOG.md              ← shared asset catalog (auto-generated)
├── CATALOG_CONFIG.yaml     ← per-project scope definitions
├── Shared/                 ← scope: global
├── Finance - Common/       ← scope: team:finance
├── [App] IT Onboarding/    ← scope: private (excluded from catalog)
└── [App] Expense Report/   ← scope: private

All of these should be protected via .workatoignore.

Project scopes

CATALOG_CONFIG.yaml

Define the visibility of each project:

# projects/CATALOG_CONFIG.yaml
projects:
  Shared:
    scope: global          # public to all teams; included in the catalog
    description: Org-wide shared logic and connections

  "Finance - Common":
    scope: team:finance    # shared within the Finance team; included in the catalog
    description: Common recipes for the Finance team

  "[App] IT Onboarding":
    scope: private         # excluded from the catalog
    description: IT onboarding for new hires

# Projects not listed here are treated as private

Scope types

Scope	Catalog inclusion	Use case
global	Every asset is listed	Shared across the org (e.g. the Shared project)
team:<name>	Every asset is listed (with team name)	Shared within a department (e.g. Finance - Common)
private	Not listed	Project-specific (e.g. a Workflow App)

Initial setup

$catalog config walks you through this interactively:

1. Show the project list under projects/.
2. Confirm the scope of each project with the user.
3. Generate CATALOG_CONFIG.yaml.

CATALOG.md and CATALOG_CONFIG.yaml are part of the kit's base .workatoignore template (templates/workatoignore.template), which $pull-project, $spec and $design place into each project — so no separate .workatoignore step is needed here.

Catalog structure

# Shared Asset Catalog
Last updated: <YYYY-MM-DD>

## Connections

| Name | Provider | Project | Scope |
|---|---|---|---|
| Shared \| Slack | slack_bot | Shared | global |
| Shared \| Jira | jira | Shared | global |
| Finance \| SAP | sap | Finance - Common | team:finance |

## Recipe Functions

### fnc: Get line manager
- **Project**: Shared (global)
- **File**: `Shared/Recipes/fnc_get_line_manager.recipe.json`
- **Input**: `employee_email` (string) — the employee's email address
- **Output**: `manager_name` (string), `manager_email` (string)
- **Purpose**: look up manager info from HRMS / Google Sheets

## MCP Servers

| Name | Project | Scope | Tools |
|---|---|---|---|
| IT Onboarding | Shared | global | 1 |

$catalog scan — procedure

1. Load scope settings

Read projects/CATALOG_CONFIG.yaml. If it's missing, point the user to $catalog config.

2. Filter target projects

Only scan projects whose scope is global or team:*. Skip private projects and projects not listed in the config.

3. Collect connections

Scan *.connection.json in the target projects. Extract name and provider from each file.

4. Collect Recipe Functions

Scan recipes whose name starts with fnc_*.recipe.json or fnc: *. From each file, extract:

• name — recipe name
• code.input.parameters_schema_json — input parameter schema (parse the JSON string)
• code.input.result_schema_json — output schema (parse the JSON string)
• The recipe's comment — purpose description

For each parameter field, extract name, type, label, optional and list them in the catalog.

5. Collect Workflow Apps / MCP servers

Scan *.lcap_app.json and *.mcp_server.json in the target projects.

6. Generate the catalog file

Write the collected info to projects/CATALOG.md. If an existing catalog is present, do a diff update (preserve manually-added descriptions like Purpose).

How other skills consume this

From $create-recipe

When $create-recipe is generating a recipe:

1. Check whether projects/CATALOG.md exists.
2. If so, load it and search for shared assets that match the requirements:
   • Connections: is there a shared connection for the same provider?
   • Recipe Functions: does the logic already exist?
3. If matches are found, propose them:

   Existing shared assets are available:
   - fnc: Get line manager (Shared / global) — manager lookup
   - Shared | Slack (slack_bot / global) — Slack connection
   
   Use these?
   

4. On user approval, generate a recipe that references the shared assets via call_recipe or config.

From $plan

In the technical-design phase of $plan <project>/<NNN>-<slug>:

1. Load the catalog.
2. From the user experience in spec.md, identify the required capabilities.
3. List the parts covered by shared assets in plan.md's ## Reused Assets.
4. Plan the remainder under ## New Components for new development.

Proposing consolidation

When you detect duplicate logic across private projects (during $learn-recipe or $plan):

• Do not expose the specific code content.
• Suggest: "the same logic appears in multiple projects; consider extracting it into a shared Recipe Function."
• The user decides whether to consolidate.
• If they do, place it in a global / team shared project and register it in the catalog.