ronkit

Ronnie's Claude Code plugin marketplace — skills, slash commands, agents, and hooks, installable in two commands.

Plugins here are organized by category, not one-plugin-per-skill: each plugin is a scope (documents, test, ...) that groups related skills. This keeps slash invocation (/plugin-name:skill-name — plugin skills are always namespaced, even with zero name collisions) meaningful instead of repeating the same word twice.

Installation

/plugin marketplace add ronniechong/ronkit
/plugin install <plugin-name>@ronkit

Plugins

Plugin	Version	Description	Install
test	0.1.1	Skills for testing and validating the ronkit marketplace itself	/plugin install test@ronkit
documents	0.9.0	Skills for managing project documentation	/plugin install documents@ronkit
security	0.2.0	Skills for information-risk security review	/plugin install security@ronkit
development	0.1.0	Skills for setting up development tooling and repo infrastructure	/plugin install development@ronkit

test

• hello — smoke-test skill. After installing, trigger it with /test:hello or by asking to "test the ronkit marketplace" — it should reply with a fixed confirmation string.

documents

• resume-later — keeps a project's living planning doc current across sessions, so clearing context between milestones never loses state. Trigger it either way:

  • Say it: mention you're wrapping up, pausing, or about to clear context. Claude may also raise it proactively when it notices a milestone just finished or the session's gotten long.
  • Run it: /documents:resume-later (type /resume and autocomplete will get you there)

  It finds your project's planning doc (asking if ambiguous), folds in anything undocumented (discoveries, decisions, milestone status), appends a concise History Log entry with a "what's next" note, flags any uncommitted git work, and asks for confirmation before it's safe to /clear.

• scaffold-project — sets up just the foundations: a private working-docs directory (full-context CLAUDE.md, STATE.md, JOURNAL.md, milestones, ritual prompts, artifacts/ for research/findings/instructions that feed project and milestone definition, issues/ for bugs and incidents found during and after development). This is a setup stage only — no code, and no public code repo, gets created here. Trigger it: /documents:scaffold-project, or ask to "start/kickstart/bootstrap a new project". Opens with a short interview (project name, purpose, working-docs location) before generating anything — it deliberately does not ask about tech stack or a public repo path, and screens answers for PII/sensitive details.

• plan-project — turns a scaffolded working-docs directory into an implementable plan through structured conversation: elaborates the idea, interrogates data sources and constraints, decides the tech stack, builds a risk register, and defines milestones as standalone deliverables with decision gates and a security review where applicable. The public code repo also gets created here, as one of the earliest milestones, once stack and scope are actually settled — not upfront at scaffold time. That milestone also asks whether the repo should get an AGENTS.md (cross-tool instructions readable by Cursor, Copilot, Gemini CLI, etc., not just Claude Code) — if yes, CLAUDE.md becomes a thin @AGENTS.md import plus Claude-specific rules, per Claude Code's own documented interop pattern. Each milestone closes only after a dedicated verification pass and the user's explicit sign-off against a plain-language summary — never on Claude's say-so alone. Trigger it: /documents:plan-project, or ask to plan, elaborate, or pressure-test a project — typically run right after scaffold-project.

• review-ai — reviews a technical document (system design, architecture, data schema, API spec — not a PRD) for AI usage, and if found, checks it against AI governance categories (data flow, vendor risk, security/guardrails, incident response, evaluation, plus escalation items like disclosure and regulatory mapping), producing a prioritized report for an engineering lead. Trigger it: /documents:review-ai, or ask to review a doc for AI governance or check for governance gaps. Read-only — never edits the source doc, gives a short summary in chat, and offers to write the full report to a file next to the doc reviewed.

security

• audit-pii — audits a repo for information risk before it goes public: exposed secrets/credentials, PII, sensitive operational details (hostnames, internal repo references), and conversational comments that leak process or names. Produces a severity-ranked report and prioritized, phased remediation (secrets/PII first, then operational refs, then comment cleanup) — never edits or commits without explicit approval. Trigger it: /security:audit-pii, or ask to check a repo for secrets, PII, or sensitive leaks before pushing or making it public.

development

• setup-doclapse — generates a .github/workflows/doc-lapse.yml for a repo adopting doc-lapse-action (a documentation-drift-detection GitHub Action). Asks for the target repo (absolute path), checks whether the Action is already configured somewhere in the repo's workflows and pauses with an abort/overwrite/write-fresh choice if so, shows doc-lapse-action's own compatibility table to inform the LLM provider/model choice, runs a two-tier config interview (essentials always, ~8 advanced inputs behind one opt-in gate), and pins the workflow to the latest published release tag (resolved live, no GitHub API/token required). Writes the file only — no git commands, no secret handling. Trigger it: /development:setup-doclapse, or ask to set up, add, or wire up doc-lapse-action for a repo.

Updating

/plugin marketplace update ronkit

then reinstall the plugin you want to refresh.

Local development

From the repo root:

claude
/plugin marketplace add .
/plugin install <plugin-name>@ronkit
# after changes:
/plugin uninstall <plugin-name>@ronkit
/plugin install <plugin-name>@ronkit