name	graph-index
description	Generate a cross-service dependency graph and compact service index from existing manifest YAML files. Activated when the prompt asks to produce graph.yaml and index.yaml.

Graph & Index Generation

You produce two derived files from a set of service manifest YAMLs.

All paths are relative to the working directory (project root):

Read the schema files. The output files must follow these schemas exactly.

When regenerating, read existing graph.yaml and index.yaml first to preserve any manually added entries (e.g., edges for dependencies not yet in individual manifests)
Include every service found in the manifests — do not skip any
For the graph: create one node per service (name, display_name, layer, repository). Create edges from dependencies.services entries — use dependency relationship, protocol, and purpose (as description). Create event_flows from events.publishes matched with events.consumes across services.
For the index: for each service extract from the manifest: name, display_name, description, layer, repository from service: section; role (array) and data_domain (array) from service: section; dependencies list with name, relationship, and protocol from dependencies.services.
Deduplicate: if the same dependency name appears more than once under a service, keep only one entry
Follow the provided schemas exactly
Write the files directly — do not output to terminal

name	graph-index
description	Generate a cross-service dependency graph and compact service index from existing manifest YAML files. Activated when the prompt asks to produce graph.yaml and index.yaml.

You produce two derived files from a set of service manifest YAMLs.

All paths are relative to the working directory (project root):

Read the schema files. The output files must follow these schemas exactly.

When regenerating, read existing graph.yaml and index.yaml first to preserve any manually added entries (e.g., edges for dependencies not yet in individual manifests)
Include every service found in the manifests — do not skip any
For the graph: create one node per service (name, display_name, layer, repository). Create edges from dependencies.services entries — use dependency relationship, protocol, and purpose (as description). Create event_flows from events.publishes matched with events.consumes across services.
For the index: for each service extract from the manifest: name, display_name, description, layer, repository from service: section; role (array) and data_domain (array) from service: section; dependencies list with name, relationship, and protocol from dependencies.services.
Deduplicate: if the same dependency name appears more than once under a service, keep only one entry
Follow the provided schemas exactly
Write the files directly — do not output to terminal

graph-index