Rhize Plugins for Claude Code: Which Ones Should You Install?

Install the plugin that solves the task you have now. Use seo-aeo-geo for website search work, obsidian-second-brain for your notes, project-launcher for planning a build and rhize-devflow for reviewing code changes. Add Perplexity Research when you need source-backed research inside the assistant, or use the decision pre-mortem skill in rhize-ops before committing to a plan. Each has different setup requirements.

The public Rhize marketplace lists 12 plugins in the catalog checked for this update. A plugin can package instructions, commands, hooks and tool configuration; it does not grant every permission or make every integration available. The comparison below is a guide to choosing a first workflow, followed by installation and verification.

How do you choose the first plugin?

Write the task as an observable result: "check why this page is not discoverable," "find the source of this client decision," or "review this change before deployment." Choose the package for that job, read its README and test the smallest useful example. Installing the entire catalog makes dependencies and overlapping responsibilities harder to diagnose.

How do you install and verify a plugin?

1. In Claude Code, run /plugin marketplace add https://github.com/Rhize-Media/rhize-plugins. This registers the catalog; it does not install every package.

2. Install the chosen package with /plugin install <plugin-name>@rhize-plugins. For example, /plugin install project-launcher@rhize-plugins installs the planning workflow. Read that plugin's prerequisites before running a connected task.

3. Restart the host and confirm that the expected skills or commands are discoverable. Configure only the accounts and capabilities you intend to use. A catalog entry or successful installation is not a live integration test.

4. Run one read-only or reversible example, inspect its evidence and check any unavailable capability. Only then authorize writes, publication or other external effects required by your real workflow.

For Codex, the current repository documents codex plugin marketplace add https://github.com/Rhize-Media/rhize-plugins and codex plugin add <plugin-name>@rhize-plugins. Verify that the particular package exposes the supported Codex entry point. Cowork and other hosts may differ in local tooling, hooks and connectors; do not infer support from Claude Code installation alone.

Website visibility: seo-aeo-geo

Choose this for site audits, keyword research and checking how pages appear in search and supported AI-answer surfaces. Live data workflows require DataForSEO credentials and may incur provider usage charges. Code-level review can address metadata, sitemaps and structured data when the relevant project is available. A report should state its queries, date and coverage; it cannot guarantee ranking or citation.

Notes and research: obsidian-second-brain

Choose this when the source material lives in your Obsidian vault. It supports cited recall, capture, research and vault-health workflows through the configured local integrations. Start with a read-only question and inspect the referenced notes. Optional qmd indexing should target approved note roots, while compiled knowledge uses separate preview-and-apply controls. Obsidian, its connector and any optional tools must be set up individually.

New projects: project-launcher

Choose this before building a new product or automation. It researches the available context, asks about unresolved requirements, develops a specification and tests the plan for gaps before scaffolding. A useful first request names the user, problem and constraints. Treat a reviewed plan as permission to begin the authorized build, not permission to deploy or spend money.

Code changes: rhize-devflow

Choose this for an existing codebase when a change needs impact analysis, testing and independent review. The current workflow is impact-map, implementation, simplify, check and review. Mutation and browser checks address cross-layer behavior. The complete semantic impact-map workflow also uses Context Manager. Passing Dev Flow review does not itself merge or deploy the change.

Long sessions: rhize-context-manager

Choose this when work spans sessions and the assistant needs explicit state, retrieval and checkpoints. It owns session start/done and context hygiene rather than production code review. Its current documentation includes optional memory and decision-support capabilities with separate configuration and evidence boundaries. Begin with the project's written state before enabling additional stores or projections.

Decisions and handoffs: rhize-ops

Choose this to pressure-test a plan or prepare a handoff. The decision pre-mortem moves through failure analysis, a verdict, a revised plan, an adversarial check and warning signals. It should distinguish facts from assumptions and does not authorize execution. Delegation workflows can prepare Jira or Slack material, but sending or creating external records requires the authorization appropriate to that workflow.

Research inside the assistant: perplexity-research

Choose this when you want Perplexity Search or Research reports, follow-ups and cited exports in a connected assistant. The local bridge keeps durable job IDs and can retrieve an existing report without submitting a new query. Setup requires macOS, the documented Python dependencies, an entitled Perplexity session and the configured authorization basis. Session transfer uses an encrypted browser-to-Keychain handoff.

This bridge uses undocumented subscription web endpoints and remains experimental. It is not the official Perplexity API. Queries can consume subscription allowance, endpoints can change, and stopping a local job does not prove remote cancellation or a refund. Keep these limits in the adoption decision instead of treating a successful test as permanent provider support.

Repeatable code: procedural-memory

Choose this when repeated work already has a verified implementation worth reusing. The plugin connects to a separately provisioned registry runtime; it does not create a working registry merely by being installed. Recall finds candidates, while trust, digest, verification and authorization checks govern execution. A matching description never grants permission to run code.

Daily planning: rhize-tasks

Choose this to turn approved Jira work into a time plan on a Mac, with calendar and reminder integrations configured for the operator. Review the work and available time before creating schedule entries. The exact account mappings and local requirements belong in the plugin's configuration, not in copied task instructions.

Client context: rhize-cowork

Choose this to establish starter context for a Cowork client: the business, voice and confirmed facts an assistant should use. Have the client owner review that context. A useful voice file is not permission to invent claims or publish a first draft.

Prospecting preparation: rhize-outreach

Choose this for evidence-backed prospect research, website concepts and outreach drafts. It separates observed website/search opportunities from assumptions about the business. Review the factual basis and recipients before any outreach; a generated draft is not a sent message.

Setup: rhize-core

Choose the setup hub when you want to discover installed packages and run their own setup steps. It can help select opt-in guard hooks and record baselines. Missing credentials or dependencies remain missing until configured and checked. The setup hub is a coordinator, not a replacement for each package's instructions.

What happened to rhize-meta?

The old rhize-meta plugin is no longer in the current marketplace. External-skill vetting and refinement moved to the separate Skill Forge tooling. If an older article tells you to install rhize-meta, use the current catalog and Skill Forge documentation instead. Do not keep two copies of a retired capability active merely to preserve an old command.

Where does Skill Forge fit?

Skill Forge is a separate tool for inspecting third-party agent skills and MCP candidates before adoption. Quarantine, scanning and review can expose issues; they cannot establish that arbitrary code or instructions are harmless in every environment. Treat scan evidence as one input to a decision, and keep the exact source and version tied to that decision.

The marketplace gives you capabilities. Your host permissions, provider accounts, repository policy and review process determine how they may be used. Browse the Rhize marketplace, choose one task and prove that workflow before adding more.

Sources and implementation references

Current catalog and installation: https://github.com/Rhize-Media/rhize-plugins

Perplexity Research setup: https://github.com/Rhize-Media/rhize-plugins/blob/main/perplexity-research/README.md