Holonic MCP Prompts
The fAI MCP server exposes six holonic/* development prompts for building
@fathym/steward extensions from inside any MCP-connected AI tool. Each scaffolds
or audits to the patterns - the create prompts read the relevant pattern guide
first, so what they generate already follows the holon shape.
holonic/create-capability
Input: Name (PascalCase), Purpose (1-2 sentences).
Generates: a CapabilityBuilder file with the .Input / .Output /
.Execute pattern plus an IDD intent block. Reads the capability patterns first.
holonic/create-agent
Input: Name (PascalCase), Purpose (1-2 sentences).
Generates: an AgentBuilder file with an AgentHandle (SendMessage /
Close) plus an IDD intent block. Reads the agent patterns first.
holonic/create-workflow
Input: Name (PascalCase), Steps (comma-separated).
Generates: a WorkflowBuilder with .Input / .State / .Operations and a
declarative .Flow() graph. Reads the workflow patterns first.
holonic/audit
Input: HolonType (Capability, Agent, Operation, Workflow, Governance,
Steward, Vault, Workbench, Mode), Code (TypeScript source).
Generates: a structural-compliance report against the F1-F7 checks (naming,
modifiers, immutability, schemas, IDD, type escape hatches, IntentRuntime typing).
Reads the relevant pattern guide first.
holonic/explain
Input: HolonType.
Generates: a full explanation of that holon type - what it is, when to use it,
key API patterns with signatures, its relationship to other holons, and a minimal
working example.
holonic/compose-steward
Input: Name (PascalCase), Description (the holons to include + their
purposes).
Generates: a holon hierarchy and a StewardModule wiring Agents, Capabilities,
Operations, Workflows, and Governance together. Reads the holonic-architecture
guide first.
These are delivered by the fAI MCP server (defined in @fathym/power-ai). For how
to connect them, see the fai MCP server docs.