Claude Code Plugins

A collection of Claude Code plugins by Filip Procházka.

Installation

Add this marketplace to Claude Code:

claude plugin marketplace add fprochazka/claude-code-plugins --scope user

Then install plugins:

claude plugin install skill-keyword-reminder@fprochazka-claude-code-plugins --scope user

Plugin dependencies

Some plugins declare dependencies in their plugin.json. Installing such a plugin installs its dependencies too. Dependencies that live in another marketplace resolve only when that marketplace is already added, so add these before you install sdlc, code-review, glab, or noisy-tools-in-subagent:

claude plugin marketplace add fprochazka/glab-discussion --scope user
claude plugin marketplace add fprochazka/glab-pipeline --scope user
claude plugin marketplace add fprochazka/bash-classify --scope user

A dependency installed this way is a Claude Code plugin (skill, hook, agent). It does not install the CLI binary the plugin wraps — glab, jq, glab-discussion, glab-pipeline, and bash-classify stay a manual install, see each plugin's README.

Plugin	Depends on
sdlc	prose, git, code-review, diagrams, glab, glab-discussion, glab-pipeline
code-review	prose, diagrams, glab, glab-discussion
glab	glab-discussion, glab-pipeline
noisy-tools-in-subagent	bash-classify-hook
agent-roles	noisy-tools-in-subagent
prose	skill-keyword-reminder

Upgrading

To upgrade installed plugins to the latest version:

claude plugin marketplace update fprochazka-claude-code-plugins
claude plugin update skill-keyword-reminder@fprochazka-claude-code-plugins

Available Plugins

Maintained here

Plugin	Description
sdlc	Software delivery workflow commands — gather context, write plans, file tickets, open MRs, babysit them to green, wrap them up, and brief next steps
code-review	Multi-agent branch code review — conventions, architecture, design craft, bugs, performance, security, release readiness, git history, and documentation reviewed in parallel
git	Git workflow skill and commit commands — vertical-slice atomic commits, the sibling/ancestor boundary test, fixups over correction commits, and dependency-ordered branch history
glab	GitLab through the glab CLI — the glab skill, MR commands for pipeline and comment state, and the mr-status review-state skill
prose	How Claude talks and writes — a reply style for chat with a per-prompt reminder hook, a rule set for docs, doc comments, commits, MR text and error messages, and a plain-language restate command
diagrams	Diagram authoring skills — Mermaid diagrams written to standalone files, rendered with mermaid-cli, and visually checked by a validator subagent before they go into a plan, an MR, or a ticket
noisy-tools-in-subagent	Forces noisy commands (builds, tests, linters, static analysis) to run inside a Sonnet subagent instead of the main context, preserving main-agent tokens
agent-roles	Two roles injected at the start of every agent — the top-level session orchestrates and delegates, every subagent works in the foreground and ends its turn with findings
skill-keyword-reminder	Automatically reminds Claude to load relevant skills when keyword triggers appear in user prompts
searxngcli	Skill for searching the web using a SearXNG instance via the searxngcli CLI
web-researcher	Iterative web research agent that searches, discovers new directions, and synthesizes findings
rabbitmqadmin	Skill for inspecting RabbitMQ instances using the rabbitmqadmin CLI (rabbitmqadmin-ng)
migrate-to-uv	Skill for migrating Python projects from Poetry, pipx, or pip to uv
gemini-cli	Skill and review agent for using Gemini CLI with massive context windows (1M tokens) for codebase analysis and second opinions

Hosted externally

Each of these is a CLI with its own Claude Code plugin in the same repository. The repository is its own marketplace, so add it with claude plugin marketplace add fprochazka/<name> --scope user before you install the plugin.

Plugin	Description
glab-discussion	Skill for working with GitLab MR discussions via the glab-discussion CLI
glab-pipeline	Agent-friendly GitLab CI pipeline inspector — dumps full pipeline state and prints a problem-driven summary
bash-classify	Auto-allows low-risk bash commands in Claude Code using bash-classify
slackcli	Skill for interacting with Slack workspaces using the slackcli CLI
metabasecli	Skill for interacting with Metabase using the metabase CLI
devin-mcp-cli	Skill for interacting with the Devin MCP server using the devin-mcp CLI
gemini-deep-research	Skill for conducting autonomous deep research using Google's Gemini Deep Research Agent
claude-code-session-tools	Commands and skill for finding and distilling past Claude Code sessions with the cc-session-tools CLI

Experiments

Plugins that explore an idea. They work, but the shape can change without notice and they get less care than the ones above.

Plugin	Description
ai-tool-use-validator	AI-powered tool use validation using LLM backends (Vertex AI, etc.) to evaluate command safety and correctness

Developing

Checks

Run make all before you push. It runs three things: make check (the repository checks in scripts/check.sh — manifests, marketplace and README consistency, frontmatter, references, hooks, shellcheck), make validate (claude plugin validate --strict over the marketplace and every plugin), and make test (the hook test suites). The same target runs in CI on every push and pull request. To debug one check, pass its id: scripts/check.sh references. A plugin's plugin.json is the source of its description: the marketplace entry and the README row below are copies of it, and make check fails when they drift apart.

Run make hooks once per clone. It points core.hooksPath at scripts/hooks, where a pre-commit and a commit-msg hook refuse anything whose staged lines or commit message contain a private identifier — a customer, host, bot or ticket prefix that has no business in a public repository. The hooks read the identifiers from a file outside the repository, named by git config hooks.forbiddenPatternsFile, because listing them here would publish the very strings they exist to keep out. Until that config key points at a real file, both hooks refuse the commit and print how to create it.

Creating a New Plugin

• Create plugins/<name>/ directory
• Create plugins/<name>/.claude-plugin/plugin.json with name, version, description
• Add plugin content (skills/, agents/, hooks/, etc.)
• Create plugins/<name>/README.md
• Add entry to .claude-plugin/marketplace.json with matching version
• Add row to "Available Plugins" table in the root README

Releasing a New Version

• Update version in plugins/<name>/.claude-plugin/plugin.json
• Update version in .claude-plugin/marketplace.json
• Commit and push

Notes

• Plugins are pinned to commit SHAs when installed
• Users must run claude plugin update <name>@fprochazka-claude-code-plugins to get updates
• Use semantic versioning: MAJOR.MINOR.PATCH

License

MIT