Skip to content

Glossary

Terms used throughout nd documentation, organized alphabetically. Terms in code font are values you type in commands or config files.

Agent (asset type)

A deployable asset that configures how a coding agent behaves. Agent assets are single Markdown files stored in the agents/ directory of a source.

Not to be confused with coding agent, which is the AI tool (like Claude Code) that consumes agent assets.

my-source/
└── agents/
    └── code-reviewer.md    # This is an agent asset

See Create asset sources for the full directory layout.

Agent config directory

The directory where a coding agent reads its configuration. For Claude Code, this is ~/.claude/ (global) or .claude/ (project). nd creates symlinks inside this directory when you deploy assets.

Configure custom agent directories in your config file using the agents[] array.

Asset

Any deployable unit that nd manages. nd supports eight asset types: skills, agents, commands, output styles, rules, context files, plugins, and hooks.

Each asset has a unique identity: the combination of its source, type, and name. Reference assets in commands using the format type/name (for example, skills/greeting).

Builtin source

A source embedded inside the nd binary. The builtin source ships nd-specific assets (skills, commands, and an agent) and is always present. It has the reserved ID builtin and the lowest priority — your sources override it when names collide.

Coding agent

An AI-powered development tool that reads assets from a config directory. Claude Code is the default coding agent. nd also supports other agents (Cursor, Windsurf, Copilot) through the agents[] config array.

Not to be confused with agent (asset type), which is a file you deploy to a coding agent.

Command

A deployable asset that defines a custom slash command for a coding agent. Command assets are single Markdown files stored in the commands/ directory of a source.

In Claude Code, deployed commands become available as /command-name in the chat interface.

Context file

A deployable asset that provides instructions or context to a coding agent. Context files have special deployment rules that differ from other asset types:

  • Global scope: deploys to the agent config directory (for example, ~/.claude/CLAUDE.md)
  • Project scope: deploys to the project root (for example, ./CLAUDE.md), not inside .claude/
  • Local files (*.local.md): deploy only at project scope

Context files use a folder-per-asset layout with an optional _meta.yaml sidecar for metadata.

Not to be confused with context window, which is the amount of text a language model processes at once.

See How nd works for deployment details.

Deploy

Create a symlink from a coding agent’s config directory to an asset in a source. Deploying does not copy files — it creates a link so that edits to the source appear instantly in the deployed location.

nd deploy skills/greeting

The opposite of deploy is remove.

See How nd works for what happens on disk.

Doctor

A comprehensive health check command that validates your entire nd setup. nd doctor checks five areas:

  1. Config file validity
  2. Source accessibility
  3. Deployment health (broken, drifted, or orphaned symlinks)
  4. Coding agent detection
  5. Git availability

Run nd doctor after editing config files or when deployments behave unexpectedly.

Export

Package assets from a source into a standalone plugin directory or marketplace listing. Exporting is the deployment path for plugins — unlike other asset types, plugins use nd export instead of symlink deployment.

nd export                    # Export a plugin
nd export marketplace        # Generate a marketplace listing

Health status

The condition of a deployed symlink. nd tracks five states:

StatusMeaning
OKSymlink exists and points to the correct target
BrokenSymlink exists but the target is missing
DriftedSymlink points to the wrong target
OrphanedThe source no longer contains this asset
MissingThe symlink was deleted externally

Check health with nd status or nd doctor. Repair issues with nd sync.

Hook

A deployable asset that defines lifecycle hooks for a coding agent. Hook assets are directories stored in the hooks/ directory of a source.

Hooks require manual registration in the coding agent’s settings.json after deployment. nd creates the symlink but does not modify settings.json.

Manifest

An optional nd-source.yaml file at the root of a source that overrides convention-based asset discovery. Use a manifest when your source does not follow nd’s standard directory layout.

When a manifest is present, nd uses it exclusively and ignores convention-based scanning.

See Create asset sources for the schema.

Operation log

A JSONL file at ~/.config/nd/logs/operations.log that records every mutating operation nd performs (deploys, removals, syncs, profile switches, and others). Each entry includes a timestamp, operation type, affected assets, and success/failure counts.

nd rotates the log automatically when it exceeds 1 MB.

Origin

How an asset was deployed. nd tracks three origins:

OriginMeaning
manualDeployed directly via nd deploy
pinnedLocked in place via nd pin
profile:<name>Deployed as part of a named profile

Origin determines what happens during a profile switch: nd keeps pinned and manually deployed assets, and swaps profile-origin assets.

Output style

A deployable asset that controls how a coding agent formats its output. Output style assets are single Markdown files stored in the output-styles/ directory of a source.

Like hooks, output styles require manual registration in the coding agent’s settings.json after deployment.

Pin

Lock a deployed asset so it persists across profile switches. nd neither removes nor redeploys pinned assets when you switch profiles.

nd pin skills/greeting      # Lock the asset
nd unpin skills/greeting    # Release the lock

Pinning changes the asset’s origin to pinned.

Plugin

A deployable asset type that uses an export workflow instead of symlink deployment. Plugin assets are directories stored in the plugins/ directory of a source.

Export plugins with nd export and generate marketplace listings with nd export marketplace. Plugins cannot be included in profiles or snapshots.

Profile

A named collection of assets that you deploy and switch between as a group. Profiles work like browser profiles for your coding agent — for example, a “work” profile with enterprise tools and a “personal” profile with hobby project assets.

nd profile create work --assets skills/jira,agents/reviewer
nd profile switch personal

nd stores profiles as YAML files in ~/.config/nd/profiles/.

See Profiles and snapshots for the full workflow.

Remove

Delete a deployed symlink, disconnecting the asset from the coding agent’s config directory. nd removes only the symlink, not the source file.

nd remove skills/greeting

The opposite of deploy.

Rule

A deployable asset that defines behavioral rules for a coding agent. Rule assets are single Markdown files stored in the rules/ directory of a source.

The coding agent auto-loads rules from its config directory. Deployment requires no manual registration.

Scope

Where nd deploys an asset. nd supports two scopes:

ScopeTarget directoryUse case
global (default)Agent config directory (~/.claude/)Assets you want in every project
projectProject config directory (.claude/)Assets specific to one project
nd deploy skills/greeting              # Global (default)
nd deploy skills/greeting --scope project  # Project

See How nd works for details.

Skill

A deployable asset that teaches a coding agent a reusable workflow or capability. Skill assets are directories containing a required SKILL.md entry-point file and optional supporting files, stored in the skills/ directory of a source.

Skills are specific to Claude Code. Other coding agents use different terminology for similar concepts (tools, instructions, prompts).

my-source/
└── skills/
    └── greeting/
        └── SKILL.md

Snapshot

A point-in-time record of all current deployments. Snapshots act as bookmarks you can restore to return to a known-good state.

nd creates auto-snapshots before destructive operations (profile switches, snapshot restores) as a safety net. nd retains the 5 most recent auto-snapshots.

nd snapshot save before-experiment
nd snapshot restore before-experiment

See Profiles and snapshots for the full workflow.

Source

A directory containing assets organized by type. nd supports three source types:

TypeDescription
localA directory on your filesystem
gitA cloned Git repository
builtinAssets embedded in the nd binary

Register sources with nd source add. nd scans sources for assets in convention-based directories (skills/, agents/, commands/, and others).

See Create asset sources for how to structure your own.

Sync

Pull the latest changes from git sources and repair broken symlinks across all deployments. nd sync fixes health status issues like broken, drifted, or orphaned symlinks.

nd sync                        # Repair all deployments
nd sync --source <source-id>   # Pull a git source and repair
nd sync --dry-run              # Preview repairs without applying

Symlink strategy

How nd constructs the symlink path when deploying. nd supports two strategies:

StrategyExampleUse case
absolute (default)~/.claude/skills/greeting -> /Users/you/my-assets/skills/greetingReadable paths for debugging
relative~/.claude/skills/greeting -> ../../my-assets/skills/greetingPortable across machines with different home directory paths

Set the default in your config file with symlink_strategy: relative or per-deploy with --relative.