Library: references

import { referenceCommand, referenceLayers } from 'buddy-agent-harness'

These exports are the reference command and the layers it reads. Use them to mount the command in your own tree, or to read reference documents from your own code with the same tiers.

The engine comes from cyber-agentkit; buddy-agent-harness re-exports it, naming itself as the plugin whose own references come first. Its documentation holds the signatures and types.

referenceLayers

function referenceLayers(options: ReferenceLayerOptions): Promise<ReferenceLayers>

Returns every layer a name is resolved through, highest precedence first: the managed tier, then the repo tier, then the user tier, then the plugin tier. Tiers lists each tier’s layers.

ReferenceLayerOptions is the one cyber-agentkit documents, except that plugin is fixed to this package, read as the plugin buddy-agent-harness, and packageRoot overrides the folder it is read from. It defaults to the installed package.

referenceCommand

const referenceCommand: cli.Command

The reference command and its subcommands, as a clibuilder command. Mount it in your own command tree to offer the same CLI. reference covers its flags and output.

The cyber-agentkit package

import { referenceLayers } from 'cyber-agentkit'
import { createReferenceCommand } from 'cyber-agentkit/commands'

A tool that resolves references without depending on buddy-agent-harness, such as a skill script that bundles the resolver, imports them from cyber-agentkit. createReferenceCommand and the report types come from cyber-agentkit/commands, not the root entry. No plugin is assumed: ReferenceLayerOptions takes plugin: { name, root }, the plugin calling the resolver, whose references/ is the first plugin layer.