Skip to content

Workflows

The marketplace-dev plugin exposes its capabilities as skills — there are no multi-phase orchestrator workflows. Each is a self-contained implementation or worker unit. Six are user-invocable slash commands (shown as /name); the four scaffolding/audit skills are agent-loaded — Claude dispatches them when the task calls for it. The Skills catalog is the canonical, generated list with full frontmatter descriptions.


Plugin scaffolding

scaffold-plugin <name> (agent-loaded)

Creates a new plugin directory with the correct, audit-clean structure: plugin.json, CLAUDE.md, install.sh, settings.json, and the standard subdirectories (agents/, skills/, hooks/, knowledge/). The generated directory passes the plugin-audit skill with zero findings on a clean install.

Use when: starting a new plugin inside a marketplace monorepo.

scaffold-marketplace <owner> (agent-loaded)

Creates a marketplace root — catalog (marketplace.json), release-please wiring, and at least one plugin slot.

Use when: establishing a new marketplace monorepo from scratch.

init-plugin-eval <name> (agent-loaded)

Scaffolds evals/<name>/{fixtures,expected}/ and a grading-contract README. An empty corpus grades as a clean no-op, so the scaffold is safe to commit before any fixtures exist.

Use when: adding the first review agent to a new or existing plugin.


Agent authoring

/agent-create

Creates new Claude Code sub-agent files following the official schema and token-efficiency budgets. Handles both review agents (JSON output, read-only tools, ≤ 40-line body) and team agents (prose output, action tools, ≤ 75-line body). Validates against the plugin-audit skill before writing. Updates the agent registry and plugin CLAUDE.md after success.

/agent-add

Convenience entry point that delegates to agent-create. Use when the user wants to add a review agent or team agent persona.

/agent-remove

Removes an agent file, cleans up all registry entries, removes cross-references, and updates documentation. Always confirms before deleting.

/agent-skill-authoring

Conventions, anti-patterns, and meta-patterns for writing skills (and the shared agent/skill philosophy). Use when creating or editing a SKILL.md file.

/agent-type-advisor <prose | file>

Recommends markdown vs script for a use-case (forward-looking) or audits an existing agent/skill file (retrospective). Cites rules R1–R10 from knowledge/agent-type-decision-rules.md.


Plugin maintenance

plugin-audit [dir] [--fix] (agent-loaded)

Generalized structural compliance check for any Claude Code plugin. Audits: agent type appropriateness, frontmatter compliance, eval coverage, and body line-count budgets. Accepts any plugin directory path; not hardcoded to one plugin. Pass --fix to apply auto-correctable findings.

/add-plugin <name@marketplace>

Installs a Claude Code plugin and registers it in the project's settings.json so the full team can replicate the install.


Summary

Skill Invocation Category What it does
scaffold-plugin agent-loaded Scaffolding New plugin skeleton
scaffold-marketplace agent-loaded Scaffolding New marketplace root
init-plugin-eval agent-loaded Scaffolding Eval fixture scaffold
agent-create /agent-create Agent authoring New agent file
agent-add /agent-add Agent authoring Add review/team agent
agent-remove /agent-remove Agent authoring Remove agent + registry
agent-skill-authoring /agent-skill-authoring Agent authoring Conventions reference
agent-type-advisor /agent-type-advisor Agent authoring Markdown vs. script recommendation
plugin-audit agent-loaded Maintenance Structural compliance check
add-plugin /add-plugin Maintenance Install + register a plugin

See the Skills catalog for full per-command descriptions and the Agents page for the single review agent this plugin ships.