convez.ai
AgenteDesenvolvimento WebIntermediário
shawn-sandy

Agentics Kit

shawn-sandy/agentics-kit

Marketplace de plugins para Claude Code com 12 extensões para desenvolvimento

Ver no GitHub
BomPluginScore · de 0 a 100

Sistema de marketplace para plugins do Claude Code que facilita descoberta, distribuição e instalação de plugins com IA. Oferece 12 plugins prontos para produção cobrindo revisão de código, planejamento, testes, workflows Git, acessibilidade e performance. Implementação modular e bem documentada para estender capacidades do Claude.

Instalação

$npx convez-ai shawn-sandy/agentics-kit

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
85%
Último commit
há 1 dia
Versão
—
Issues abertas
0
Contribuidores
2

Casos de uso

  • Revisão automática de código
  • Automação de workflows Git
  • Testes e QA
  • Planejamento de sprints
  • Análise de acessibilidade
  • Desenvolvimento de novos plugins

Veredito

Kit maduro (v4.0) com marketplace bem estruturado e 12 plugins ativos. Decisão clara de remover 6 plugins mantém qualidade alta. Documentação excelente e CI/CD robusto indicam manutenção ativa — recomendado para times que usam Claude Code.

README

A tradução para o português deste README ainda está na fila.

agentics

Publish dist

A marketplace system for Claude Code plugins — enabling discovery, distribution, and installation of AI-powered plugins that extend Claude's capabilities across code review, planning, testing, git workflows, accessibility, and more.

Marketplace: agentics-kit v4.0.0 · 12 plugins · Requires Claude Code 1.0.33+ · View all plugins · Browse docs

Breaking change — v4.0.0: Six plugins have been removed from the marketplace: agent-creator, agent-reviewer, agentic-plugin-dev, code-simplifier, marketplace-builder, and react-perf-analyzer. Their source directories are retained in the repository but are no longer installable via the marketplace. See CHANGELOG.md for details.


Table of Contents


Quick Start

Install from the marketplace (no cloning required):

# 1. Register the agentics-kit marketplace in Claude Code
/plugin marketplace add shawn-sandy/agentics-kit

# 2. Install the plugins you want
/plugin install code-review@agentics-kit
/plugin install git-agent@agentics-kit
/plugin install memory-tools@agentics-kit

Or load any plugin directly for local testing:

git clone https://github.com/shawn-sandy/agentics-kit.git
cd agentics-kit
claude --plugin-dir ./kit/plugins/code-review
# Then ask: "Review this code for issues"

Overview

The agentics project serves two purposes:

PurposeWhat it contains
Active Plugins12 marketplace plugins in kit/plugins/ — installable via /plugin install, covering code review, planning, testing, git workflows, accessibility, and more
Archived Plugins6 removed plugins with source directories retained in kit/plugins/ — loadable locally via --plugin-dir but not available via /plugin install
Marketplace Infrastructureagentics-kit marketplace manifest (marketplace.json) that enables installation via /plugin install

Every plugin in this repo is a working, production-quality tool you can install and use immediately.


Prerequisites

Required

  • Claude Code CLI version 1.0.33 or later
# Verify version
claude --version

Install from the official docs if not present.

Optional

  • Git — for cloning the repository locally
  • GitHub CLI (gh) — used by the git-agent plugin for GitHub PR creation
  • GitLab CLI (glab) — used by agent-ship for GitLab MR creation
  • jq — JSON processor, useful for debugging plugin manifests

Platform Support

PlatformSupport
macOS 12.0+Full support
Linux (Ubuntu 20.04+)Full support
Windows (WSL2)Recommended over native Windows

Project Structure

agentics/
├── .claude-plugin/
│   └── marketplace.json          # Marketplace manifest — the agentics-kit registry
├── .claude/
│   ├── rules/                    # Scoped authoring rules (plugin patterns, marketplace, testing)
│   └── settings.json             # Project-level Claude Code settings and hooks
├── kit/
│   └── plugins/                  # 12 plugins in marketplace (6 archived directories retained)
│       ├── agent-creator/
│       ├── agent-reviewer/
│       ├── agentic-plugin-dev/
│       ├── code-review/
│       ├── code-simplifier/
│       ├── code-testing-agent/
│       ├── git-agent/
│       ├── issue-agent/
│       ├── marketplace-builder/
│       ├── memory-tools/
│       ├── plan-agent/
│       ├── plan-interview/
│       ├── product-plans/
│       ├── react-perf-analyzer/
│       ├── settings-sync/
│       ├── skill-reviewer/
│       ├── social-media-tools/
│       └── wcag-compliance-reviewer/
├── tests/
│   └── fixtures/                 # Validation test fixtures
├── docs/
│   ├── index.html                # Landing hub — links to Plans gallery & Social cards
│   ├── plans/                    # HTML plan files & filterable gallery (index.html)
│   ├── media/social/             # Social media card gallery
│   └── guides/                   # Reference guides (auto-load setup, etc.)
├── examples/                     # Demo scripts
├── CONTRIBUTING.md
├── CHANGELOG.md
└── README.md

Installation

Install via Marketplace (Recommended)

The marketplace approach uses sparse cloning — only the plugin you install is fetched, not the entire repository.

Step 1: Register the marketplace

/plugin marketplace add shawn-sandy/agentics-kit

Step 2: Install individual plugins

/plugin install code-review@agentics-kit
/plugin install plan-interview@agentics-kit
/plugin install memory-tools@agentics-kit
/plugin install git-agent@agentics-kit
/plugin install skill-reviewer@agentics-kit
/plugin install code-testing-agent@agentics-kit
/plugin install wcag-compliance-reviewer@agentics-kit
/plugin install product-plans@agentics-kit
/plugin install plan-agent@agentics-kit
/plugin install settings-sync@agentics-kit
/plugin install social-media-tools@agentics-kit
/plugin install issue-agent@agentics-kit

Or install all at once — paste the full block above into your Claude Code session.

Auto-enable for Your Team (Declarative Config)

Instead of every contributor running /plugin marketplace add and /plugin install for each plugin, you can declare the kit once in settings.json using two keys: extraKnownMarketplaces (makes Claude Code aware of the marketplace) and enabledPlugins (marks which plugins should be enabled by default).

This repo ships these keys in project scope (.claude/settings.json). On first use, Claude Code prompts the user to trust the repo and add the marketplace; once added, the listed plugins are enabled by default — no per-plugin install commands. To enable the same kit across all your other repos, add the two keys to your personal user settings (~/.claude/settings.json):

{
  "extraKnownMarketplaces": {
    "agentics-kit": {
      "source": { "source": "github", "repo": "shawn-sandy/agentics-kit" }
    }
  },
  "enabledPlugins": {
    "memory-tools@agentics-kit": true,
    "code-review@agentics-kit": true,
    "plan-interview@agentics-kit": true,
    "wcag-compliance-reviewer@agentics-kit": true,
    "skill-reviewer@agentics-kit": true,
    "code-testing-agent@agentics-kit": true,
    "git-agent@agentics-kit": true,
    "product-plans@agentics-kit": true,
    "settings-sync@agentics-kit": true,
    "social-media-tools@agentics-kit": true,
    "plan-agent@agentics-kit": true,
    "issue-agent@agentics-kit": true
  }
}

This config lives in settings.json, not CLAUDE.md. enabledPlugins is an object ("name@agentics-kit": true), not an array. Merge these keys into any existing settings rather than overwriting the file.

Heads-up on first run and web sessions: the marketplace add is gated by a one-time trust/consent prompt. In a non-interactive context (such as a fresh Claude Code on the web session that can't answer prompts or run /plugin), the kit may not load until that prompt is accepted. If a session doesn't pick up the plugins, accept the trust prompt or run /plugin marketplace add shawn-sandy/agentics-kit once.

See the full team setup guide — including scope choices and caveats — in docs/plugin-auto-load-setup.md.

Load Locally for Testing

Clone the repo and load any plugin directly with --plugin-dir:

git clone https://github.com/shawn-sandy/agentics-kit.git
cd agentics-kit

# Load a single plugin (starts an interactive Claude session)
claude --plugin-dir ./kit/plugins/code-review

# Load multiple plugins simultaneously
claude --plugin-dir ./kit/plugins/code-review \
       --plugin-dir ./kit/plugins/plan-interview \
       --plugin-dir ./kit/plugins/git-agent

# Or run with a prompt directly (non-interactive)
claude --plugin-dir ./kit/plugins/code-review "Review this file for bugs"

Troubleshooting

claude: command not found

Claude Code CLI is not installed or not in your PATH.

Plugin not loading

The path may be wrong or the manifest may be missing.

# Verify the plugin directory exists and has a manifest
ls -la ./kit/plugins/code-review/.claude-plugin/plugin.json

# Use an absolute path if relative paths don't work
claude --plugin-dir /full/path/to/agentics/kit/plugins/code-review

Permission errors

ls -la ./kit/plugins/code-review
chmod -R +r ./kit/plugins/code-review

"Input must be provided" error with --plugin-dir

This occurs when Claude Code can't start the interactive session. Try:

# Provide a prompt directly
claude --plugin-dir ./kit/plugins/code-review "List available commands"

# Or pipe stdin
echo "Review this code" | claude --plugin-dir ./kit/plugins/code-review

# Verify manifest is valid JSON
cat kit/plugins/code-review/.claude-plugin/plugin.json | jq

Usage Guide

Commands vs Skills vs Agents vs Hooks

Plugins can include four types of components:

TypeInvocationExampleUse When
CommandsExplicit: /plugin:name/plan-interview:deep-grill plan.mdUser controls exactly when it runs
SkillsAutomatic: matches your intent"Review this code for bugs"Claude detects the need from conversation
AgentsDelegated: spawned as subprocessesBackground git commitWork should run without blocking
HooksEvent-driven: lifecycle triggersPre-commit filename validationActions must happen automatically

Most plugins use skills (automatic activation). Ask naturally and Claude activates the right skill.

Single Plugin

# Start an interactive session with a plugin loaded
claude --plugin-dir ./kit/plugins/git-agent

# Inside the session, use naturally:
# "Commit my changes with a conventional message"
# "Create a PR for this branch"
# /git-agent:ship-bg

Multiple Plugins

claude --plugin-dir ./kit/plugins/code-review \
       --plugin-dir ./kit/plugins/git-agent \
       --plugin-dir ./kit/plugins/plan-interview

# All skills and commands from all three plugins are available

Use /help inside any Claude session to list all active commands.


Plugins

Code Quality


code-review v3.3.2

Systematic code review across quality, bugs, security, and best practices with severity-ranked findings, actionable feedback, and line numbers.

Commands:

CommandDescription
/code-review:fix-branchReview all branch changes vs the default branch, then autonomously fix blocking, major, and minor issues until the branch is clean. Refuses on a dirty working tree. Leaves fixes uncommitted.

Skills (activate automatically):

SkillActivates when you ask to...
code-review-agentReview code, check files for problems, look over a PR or diff, assess quality or complexity, find bugs or security issues, detect breaking changes, or evaluate regression risk

Agents:

AgentPurpose
agent-code-reviewerInternal background code review agent for delegation from other agents or automated workflows
claude --plugin-dir ./kit/plugins/code-review
# "Review this function for security issues"
# "Check for breaking changes in my last commit"
# /code-review:fix-branch

View Documentation


Testing


code-testing-agent v3.4.4

Analyze code and suggest specific, purpose-driven tests tied to actual behavior and intent — not arbitrary coverage.

Skills:

SkillActivates when you ask to...
code-testing-agentSuggest tests, identify what to test, or find untested behavior
reviewing-testsReview tests, audit test quality, or improve a test suite
running-testsRun tests, check if tests pass, or verify changes
tdd-fixTDD-fix a bug or run a red-green cycle — manual invoke only
tdd-loopTDD a new feature or write tests first — manual invoke only
claude --plugin-dir ./kit/plugins/code-testing-agent
# "What tests should I write for this function?"
# "Review my test suite for gaps"
# "TDD fix this bug — write a failing test then make it green"

View Documentation


Planning


plan-interview v2.2.7

Stress-test implementation plans with structured multi-round interviews before coding begins. Auto-routes product plans to the panel review skill and always emits an interview HTML artifact.

Commands:

CommandDescription
/plan-interview:plan-interview [plan-file-path]Stress-test a plan with a structured interview across technical, UX, edge case, and out-of-scope domains
/plan-interview:deep-grill [plan-file-path]Walk each decision branch in an implementation plan with focused questions and codebase exploration
/plan-interview:plan-status [plan-file-path]Check and update the lifecycle status of a plan file (todo, in-progress, completed) with type classification
/plan-interview:update-plan-status [directory-path] [--force]Process multiple plan files in a directory — analyze codebase evidence and add/update YAML frontmatter in bulk
/plan-interview:plan-hygiene [directory-path]Scan plan directories for randomly-named files and rename them to descriptive kebab-case names
/plan-interview:review-rename-plans [plan-file-or-directory]Review plan filenames and offer to rename files whose names don't match their intent
/plan-interview:documenting-plans [plan-file-path]Generate developer-friendly documentation at docs/.md from a completed plan file
/plan-interview:markdown-to-html [file-path] [--theme=default|developer|document|minimal] [--mode=auto|plan|doc] [--background] [--no-open]Convert a markdown file or plan to a rich, self-contained HTML document viewable in any browser
/plan-interview:plan-maintenance [--archive] [--index] [--variants] [--all] [--background]Archive completed plans as HTML, generate a README index, and review variant/duplicate files
/plan-interview:plan-to-html [plan-file-path]Deprecated — use /plan-interview:markdown-to-html instead

Skills:

SkillActivates when you ask to...
plan-interviewStress-test or validate a technical plan
deep-grillDeep grill or stress-test a plan — manual invoke only
documenting-plansDocument a completed plan — manual invoke only
markdown-to-htmlConvert a markdown file or plan to HTML
plan-statusCheck or update a plan's status
plan-to-htmlConvert a plan to HTML

Agents:

AgentPurpose
plan-documenterBatch documentation agent that scans the plans directory for completed plans without corresponding documentation in docs/, then invokes the documenting-plans skill for each one
claude --plugin-dir ./kit/plugins/plan-interview
# /plan-interview:plan-interview docs/plans/my-plan.md
# /plan-interview:deep-grill docs/plans/my-plan.md
# "Stress-test this plan"

View Documentation


product-plans v3.4.9

Improve, optimize, and update product plans, PRDs, and feature proposals using a cross-functional Agent Team — PM, Lead Developer, UX Designer, Frontend Engineer, Accessibility Expert, and Security Expert. Produces a 15-section consolidated report, applies improvements to the source plan, and appends findings to any existing plan-interview HTML artifact.

Commands:

CommandDescription
/product-plans:product-plans-bg <path>Run the product-plans review panel in the background

Skills (activate automatically):

SkillActivates when you ask to...
plan-review-agentsReview or improve a product plan, PRD, or feature proposal with comprehensive PM, Dev, UX, Frontend, A11y, and Security coverage

Agents:

AgentPurpose
agent-product-plansBackground product-plan panel agent — runs the full six-reviewer cross-functional panel without blocking the parent session
product-reviewer-pmProduct Manager reviewer teammate
product-reviewer-lead-developerLead Developer reviewer teammate
product-reviewer-ux-designerUX Designer reviewer teammate
product-reviewer-frontend-engineerLead Frontend Engineer reviewer teammate
product-reviewer-accessibility-expertAccessibility Expert reviewer teammate
product-reviewer-security-expertSecurity Expert reviewer teammate
claude --plugin-dir ./kit/plugins/product-plans
# "Review this PRD with your full panel"
# "What would the security reviewer say about this plan?"
# /product-plans:product-plans-bg docs/plans/my-feature.md

View Documentation


plan-agent v1.10.1

Plan creation and review on demand or via ambient activation. Run /plan-agent:implementation-plan <objective> for the full Steps 0–8 planning workflow with built-in structured interview, an end-to-end self-verification gate, and a mandatory acceptance-criteria gate during implementation. Spawn a seven-reviewer Agent Team with /plan-agent:review-plan, finalize and mark plans completed with /plan-agent:finalize-plan, or generate Anthropic-best-practice AI prompts with /plan-agent:refine-prompt. Accepts GitHub/GitLab issue URLs and #n references to auto-seed plans. Generates self-contained interactive HTML plans with copy-paste implement prompts and optional workflow prompts for complex plans. A PostToolUse hook auto-regenerates the plans gallery index; a filename hook enforces verb-target kebab-case.

Commands:

CommandDescription
/plan-agent:review-plan-bg <path>Run the seven-reviewer plan-review Agent Team in the background — validates the path, spawns agent-review-plan, and returns an ack immediately

Skills:

SkillActivates when you ask to...
implementation-planCreate a plan via /plan-agent:implementation-plan <objective> — also auto-activates on plan-document intent
review-planSpawn a seven-reviewer Agent Team (architecture, completeness, testability, risk, conventions, + UI-conditional UX and accessibility) to review a plan, synthesize findings, and apply improvements in place
finalize-planReview a plan for completion evidence with per-criterion verification and mark it completed — manual invoke only
refine-promptGenerate a copy-pasteable AI prompt grounded in Anthropic best practices (role, XML structure, CoT, examples) — command only
plans-libraryBrowse plans, view plan history, or open the plans index
plans-openReopen the plans gallery without rebuilding

Agents:

AgentPurpose
agent-review-planBackground plan-review agent — invokes the review-plan skill with --background and reports the updated path on completion
plan-reviewer-architecture · -completeness · -testability · -risk · -conventionsFive core reviewer teammates, always spawned by the Agent Team
plan-reviewer-ux · -accessibilityTwo UI-conditional reviewer teammates, spawned when UI signals are detected

The review-plan Agent Team requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 and Claude Code ≥ 2.1.32.

Hooks:

HookTriggerPurpose
validate-plan-filenamePostToolUse (Write/Edit)Enforces verb-target kebab-case filenames on plan files
rebuild-plans-indexPostToolUse (Write/Edit/MultiEdit)Auto-regenerates the gallery index when plans change
claude --plugin-dir ./kit/plugins/plan-agent
# /plan-agent:implementation-plan "Add dark mode support to the settings page"
# /plan-agent:implementation-plan https://github.com/org/repo/issues/42
# /plan-agent:review-plan docs/plans/add-dark-mode-toggle.html
# /plan-agent:review-plan-bg docs/plans/add-dark-mode-toggle.html
# /plan-agent:finalize-plan add-dark-mode-toggle.html
# /plan-agent:refine-prompt
# "Browse my plans"

View Documentation


Git & Workflow


git-agent v3.10.6

Automated git workflow — create branches, commit with conventional messages, and create PRs. Auto-links plan issue references in PR descriptions.

Commands:

CommandDescription
/git-agent:commit-bgFire off the agent-commit subagent in the background to stage and commit the working tree, then return control immediately
/git-agent:pr-bgFire off the agent-pr subagent in the background to push the current branch and open a GitHub PR, then return control immediately
/git-agent:ship-bgFire off the agent-ship subagent in the background to commit, push, and open a PR/MR end-to-end, then return control immediately

Skills:

SkillActivates when you ask to...
branch-agentCreate or start a new branch — manual invoke only
commit-agentCommit or save work to git — manual invoke only
pr-agentCreate a PR or open a pull request — manual invoke only
shipShip changes or commit and create a PR — manual invoke only
ship-autonomousAutonomously ship or watch CI — runs the full ship pipeline with CI polling and bounded autofix

Agents:

AgentPurpose
agent-commitBackground git commit agent — stages all working-tree changes and creates a conventional commit message without user interaction
agent-prBackground pull-request creation agent — pushes the current branch if needed and opens a GitHub pull request with an auto-generated summary
agent-shipBackground end-to-end ship agent — stages, commits, pushes, and opens a pull/merge request in one autonomous flow
claude --plugin-dir ./kit/plugins/git-agent
# "Commit my changes"
# "Create a branch for this feature"
# "Ship it"
# /git-agent:ship-bg

View Documentation


settings-sync v1.0.2

Back up and restore Claude Code user settings to a dedicated git repo. Routine-compatible for automated backups.

Skills (activate automatically):

SkillActivates when you ask to...
settings-backupBack up, save, or sync Claude Code settings to a git repo — also activates for routine-scheduled setting backups
settings-restoreRestore, import, or recover Claude Code settings from a backup git repo
claude --plugin-dir ./kit/plugins/settings-sync
# "Back up my Claude Code settings"
# "Restore my settings from my backup repo"

View Documentation


Accessibility & Performance


wcag-compliance-reviewer v1.2.3

Review HTML/CSS and React/TypeScript code for WCAG 2.2 Level AA accessibility compliance.

Skills (activate automatically):

SkillActivates when you ask to...
wcag-compliance-reviewerCheck WCAG compliance, audit accessibility, review HTML/CSS and React code for WCAG 2.2 Level AA violations, or identify and fix accessibility issues
claude --plugin-dir ./kit/plugins/wcag-compliance-reviewer
# "Check this component for accessibility issues"
# "Is this page WCAG 2.2 AA compliant?"
# "Audit the ARIA usage in this form"

View Documentation


skill-reviewer v2.2.6

Review and plan Claude Code skills, and run tests for changed files — audit SKILL.md files, scaffold new skills, and verify test coverage.

Commands:

CommandDescription
/skill-reviewer:check-descriptionMeasure description-frontmatter length for one or more SKILL.md files and warn if any exceed the 200-char budget

Skills:

SkillActivates when you ask to...
reviewing-skillsReview, audit, or score a skill
planning-skillsPlan or scaffold a new skill
auditing-allowed-toolsAudit, fix, or review tool permissions
optimizing-skill-frontmatterOptimize SKILL.md frontmatter — manual invoke only
claude --plugin-dir ./kit/plugins/skill-reviewer
# "Review this SKILL.md file"
# "Help me plan a new skill"
# "What allowed-tools should this skill have?"

View Documentation


Productivity


memory-tools v3.1.3

Audit and optimize CLAUDE.md project memory files against Claude Code best practices.

Skills (activate automatically):

SkillActivates when you ask to...
agentic-memory-doctorAudit, optimize, or diagnose a CLAUDE.md or project memory file — also activates when Claude appears to be ignoring project instructions
path-rules-advisorCreate path-specific rules, add rules for file types or directories, or organize Claude rules in .claude/rules/
claude --plugin-dir ./kit/plugins/memory-tools
# "Audit my CLAUDE.md file"
# "Claude keeps ignoring my instructions — what's wrong?"
# "Create path-specific rules for my src/ directory"

View Documentation


social-media-tools v2.10.1

Discover teachable code, blog posts, videos, GitHub snippets, and selected/pasted code from git history or a codebase path, scrub for secrets, draft instructional platform-aware copy with concrete takeaways, and generate styled dark-mode social cards (1024px wide) for LinkedIn, Twitter/X, Bluesky, and Substack. Generate a SOCIAL.md project config for default platform, tone, and content preferences.

Commands:

CommandDescription
/social-media-tools:digestScan recent git history or a codebase path for shareable code, scrub for secrets, and draft code-share prompts

Skills:

SkillActivates when you ask to...
social-shareShare what you're working on or post code, a blog, video, or project update
share-githubShare a code snippet from a GitHub repository
share-sessionShare your session, session recap, what you worked on today, or a session summary
security-scrubCheck for secrets or review a diff for leaks
share-selectionShare, post, or tweet selected, highlighted, or pasted code
media-libraryBrowse the media library or find a prior post
share-scanFind commits worth sharing or create a digest
share-blogShare a blog post or article on social media
share-codePost or share a code change
share-explanationExplain how a project file, component, or concept works — "how does X work" or "explain X"
share-videoShare or promote a video on social media
share-projectAnnounce features, bugs, changes, or releases on social media — manual invoke only
share-initSet up social sharing preferences and generate a SOCIAL.md project config
claude --plugin-dir ./kit/plugins/social-media-tools
# "Create a social card for this feature"
# "Draft a LinkedIn post about my latest commit"
# /social-media-tools:digest

View Documentation


issue-agent v0.2.4

Create GitHub and GitLab issues from any context — selection, session, bug, or feature — with host auto-detection, a confirmation gate before writing, and automatic browser open on creation.

Skills (activate explicitly):

SkillActivates when you ask to...
create-issueFile, open, or create an issue or ticket — manual invoke only
claude --plugin-dir ./kit/plugins/issue-agent
# /issue-agent:create-issue bug "Login button throws 500 on mobile"
# /issue-agent:create-issue feature "Add dark mode toggle to settings page"
# /issue-agent:create-issue session

View Documentation


Plugin Reference Table

PluginVersionCategoryComponents
code-review3.3.2development1 command, 1 skill, 1 agent
code-testing-agent3.4.4testing5 skills
plan-interview2.2.7development10 commands, 6 skills, 1 agent, 1 hook
product-plans3.4.9productivity1 command, 1 skill, 7 agents
plan-agent1.10.1productivity1 command, 6 skills, 8 agents, 2 hooks
git-agent3.10.6development3 commands, 5 skills, 3 agents
settings-sync1.0.2productivity2 skills
wcag-compliance-reviewer1.2.3security1 skill
skill-reviewer2.2.6development1 command, 4 skills, 1 hook
memory-tools3.1.3development2 skills
social-media-tools2.10.1productivity1 command, 13 skills
issue-agent0.2.4development1 skill

Removed Plugins

The following plugins have been removed from the agentics-kit marketplace as of v4.0.0. They will not appear when browsing or installing from the marketplace — /plugin install will not find them.

Their source directories are retained in the repository as reference implementations. You can still load any of them locally with --plugin-dir:

git clone https://github.com/shawn-sandy/agentics-kit.git
cd agentics-kit
claude --plugin-dir ./kit/plugins/<plugin-name>
PluginLast VersionRemovedReasonReplacement
agent-creator1.1.22026-05-29Redundant with agentic-plugin-devUse agentic-plugin-dev (also archived; see below)
agent-reviewer1.0.22026-05-29Overlaps with skill-reviewer/plugin install skill-reviewer@agentics-kit
marketplace-builder1.1.22026-05-29Redundant with agentic-plugin-devUse agentic-plugin-dev (also archived; see below)
react-perf-analyzer1.3.12026-05-29Too specialized for React-only projects/plugin install code-review@agentics-kit
agentic-plugin-dev1.2.22026-05-29Functionality consolidated into existing skills/plugin install skill-reviewer@agentics-kit
code-simplifier1.0.22026-05-29Structural analysis covered by code-review/plugin install code-review@agentics-kit

To load a removed plugin locally:

# Example — load agent-creator for local use
claude --plugin-dir ./kit/plugins/agent-creator

# Load multiple, mixing active marketplace plugins with archived ones
claude --plugin-dir ./kit/plugins/code-review \
       --plugin-dir ./kit/plugins/agent-creator

Re-registering any of these plugins in marketplace.json requires explicit confirmation — see the Removed Plugins registry for the removal rationale before proceeding.


Contributing

Reporting Bugs

Open a GitHub Issue with:

  • Plugin name and version
  • Claude Code CLI version (claude --version)
  • Steps to reproduce
  • Expected vs actual behavior
  • Error messages or screenshots

Creating a New Plugin

Step 1: Scaffold the directory structure

kit/plugins/my-plugin/
├── .claude-plugin/
│   └── plugin.json          # Required: name, description (NO version field)
├── commands/                 # Slash commands — optional
│   └── my-command.md
├── skills/                   # Auto-activated skills — optional
│   └── my-skill/
│       └── SKILL.md
├── agents/                   # Background subagent definitions — optional
│   └── my-agent.md
└── README.md

Step 2: Create the plugin manifest

{
  "name": "my-plugin",
  "description": "What this plugin does"
}

Version rule: version belongs only in .claude-plugin/marketplace.json. Adding it to plugin.json silently overrides the marketplace version and causes conflicts.

Step 3: Test locally

claude --plugin-dir ./kit/plugins/my-plugin

# Inside the session, test your commands/skills:
# /my-plugin:my-command
# "Invoke my-plugin on this file"

Step 4: Register in the marketplace

Add an entry to .claude-plugin/marketplace.json:

{
  "name": "my-plugin",
  "source": {
    "source": "git-subdir",
    "url": "https://github.com/shawn-sandy/agentics.git",
    "path": "kit/plugins/my-plugin"
  },
  "version": "1.0.0",
  "description": "What this plugin does",
  "category": "development",
  "tags": ["specific", "relevant", "tags"]
}

Available categories: development · testing · productivity · security · documentation · learning

Pull Request Process

  1. Create a feature branch from main
  2. Build and test your plugin locally with --plugin-dir
  3. Bump the plugin's version in marketplace.json (must be higher than main)
  4. Include the relevant plan file from docs/plans/ in your commit
  5. Submit a PR with a clear description

PR Checklist:

  • plugin.json has name and description — no version field
  • Version bumped in marketplace.json and higher than version on main
  • Plugin tested locally with claude --plugin-dir
  • README.md included in plugin directory
  • Homepage URL points to the plugin directory: https://github.com/shawn-sandy/agentics/tree/main/kit/plugins/my-plugin
  • CHANGELOG.md updated (for existing plugins)
  • Plan file committed alongside changes

Versioning

BumpWhen
PATCH x.y.ZBug fix, typo, metadata correction
MINOR x.Y.zNew command, skill, agent, or hook added
MAJOR X.y.zRemoving/renaming a command/skill/agent, changing argument format

Commit message conventions:

fix(kit/plugins/my-plugin): bump version to 1.0.1     # patch
feat(kit/plugins/my-plugin): bump version to 1.1.0    # minor
feat(kit/plugins/my-plugin)!: bump version to 2.0.0   # major

See CONTRIBUTING.md for the full guide.


Development

Authoring Rules

Detailed patterns live in .claude/rules/:

FileScopeContent
plugin-patterns.mdkit/plugins/**Command/skill patterns, progressive disclosure, pitfalls
marketplace.mdglobalCategories, tagging, versioning, registration
testing.mdtests/**Test fixture guidelines
plan-hygiene.md**/plans/**Pre-commit plan file rename checks
skill-authoring.mdglobalSkill description format, allowed-tools, trigger phrases

Project Hooks (Auto-Active)

Three hooks run automatically after every file Write/Edit:

  1. JSON validation — validates marketplace.json syntax
  2. Uncommitted plan warning — alerts if plan files are unstaged alongside plugin changes
  3. Version guard — checks that the plugin version was bumped when marketplace.json was modified

Running Tests

# Run the demo test suite
bash tests/demo/run.sh

# Run the docs/pages smoke tests
bash tests/pages/test-docs-hub.sh
bash tests/pages/test-root-redirect.sh

# Check fixture validity
ls tests/fixtures/valid-plugin/

Browsing Docs Online

The docs landing hub is deployed to GitHub Pages at shawn-sandy.github.io/agentics and links to the Plans gallery and Social Media card gallery.

Browsing Docs Locally

Two ways to serve the HTML galleries (docs/plans/ and docs/media/social/) locally. The root docs/index.html is a landing hub with cards linking to each gallery:

Script — auto-selects a free port:

bash scripts/serve-docs.sh
# Plans gallery:  http://localhost:<port>/plans/
# Media library:  http://localhost:<port>/media/social/

Pass an explicit port as the first argument to pin it:

bash scripts/serve-docs.sh 8900

Launch configs — fixed ports (.claude/launch.json):

ConfigURLDirectory
plans-galleryhttp://localhost:8901docs/plans/
media-libraryhttp://localhost:8902docs/media/social/
docs-allhttp://localhost:8900docs/ (landing hub at root)

Distribution

Plugin releases are published daily to shawn-sandy/agentics-kit — a clean, plugin-only distribution repository containing only the installable plugin files with no development scaffolding, test fixtures, or internal tooling.

Install from the distribution repo:

/plugin marketplace add shawn-sandy/agentics-kit

Repository roles:

RepoPurpose
shawn-sandy/agenticsDevelopment workspace — full source, tests, plans, CI, and authoring tooling
shawn-sandy/agentics-kitClean distribution — plugin files only, built and published by the daily publish workflow

Publish workflow: publish-dist.yml — runs daily and pushes the current kit/plugins/ contents to shawn-sandy/agentics-kit.

Trigger a manual publish:

gh workflow run publish-dist.yml --repo shawn-sandy/agentics

CI/CD

GitHub Actions workflows:

WorkflowTriggerPurpose
claude.yml@claude mention in issues/PRsRespond to questions and implement requested changes
claude-code-review.ymlPR opened / synchronized / reopenedAutomated code review as PR comments
deploy-pages.ymlPush to main (docs changes)Deploy docs/ to GitHub Pages
update-readme.ymlEvery Sunday at 00:00 UTCKeep the README in sync with marketplace.json

To trigger Claude in any issue or PR comment, mention @claude:

@claude Can you review the skill descriptions in this plugin?

Resources

ResourceLink
Claude Code Docshttps://code.claude.com/docs/en
Plugin Creation Guidehttps://code.claude.com/docs/en/plugins
Plugin Referencehttps://code.claude.com/docs/en/plugins-reference
Plugin Marketplaceshttps://code.claude.com/docs/en/plugin-marketplaces
Discover Pluginshttps://code.claude.com/docs/en/discover-plugins
ChangelogCHANGELOG.md
RoadmapROADMAP.md
Security PolicySECURITY.md
Code of ConductCODE_OF_CONDUCT.md

License

MIT License — Copyright (c) 2026 Shawn Sandy

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Ver o original no GitHub

Bom para

Relacionados