<p align="center"> <img src="assets/fakoli-banner.png" alt="Fakoli Plugins Marketplace" width="100%"> </p>
<p align="center"> <a href=".github/workflows/validate.yml"><img src="https://github.com/fakoli/fakoli-plugins/actions/workflows/validate.yml/badge.svg" alt="CI"></a> <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-blue.svg" alt="MIT License"></a> <img src="https://img.shields.io/github/stars/fakoli/fakoli-plugins?style=social" alt="GitHub Stars"> </p>
<h1 align="center">Fakoli Plugins Marketplace</h1>
<p align="center"><strong>Extend Claude Code with production-grade plugins.</strong></p>
<p align="center"> A curated collection of battle-tested Claude Code plugins — covering Google Workspace automation, AI image generation, text-to-speech, diagram authoring, secure web fetching, and marketplace self-management. The active catalog combines end-user tools, development workflows, durable project state, and CI-validated marketplace practices. </p>
---
Quick Start
Add this marketplace to Claude Code with one command:
/plugin marketplace add fakoli/fakoli-plugins
Then install any plugin:
/plugin install gws
/plugin install safe-fetch
/plugin install nano-banana-pro
---
What are Claude Code Plugins?
Claude Code plugins extend the assistant with domain-specific capabilities. A plugin can bundle:
- Skills — Reusable context files that teach Claude how to use a tool or follow a workflow
- Commands — Slash commands (e.g.
/send-email,/speak,/fetch) that invoke specific behaviors - Agents — Isolated sub-agents for complex, multi-step operations
- Hooks — PreToolUse / PostToolUse interceptors that modify or guard Claude's actions
Plugins live in directories with a .claude-plugin/plugin.json manifest. The marketplace validates every plugin against a JSON Schema on every push and pull request.
---
The Fakoli Ecosystem
Fakoli is a coherent plugin marketplace rather than a loose grab bag. It has four layers:
- Work orchestration —
fakoli-flow,fakoli-crew, andfakoli-state
define how work is planned, staffed, claimed, verified, and recorded.
- Infrastructure judgment —
systems-thinkinghelps teams examine hidden
costs, dependencies, caveats, and risks before committing to large changes.
- Plugin production practices —
marketplace-manager,cli-to-plugin,
and fakoli-plugin-critic help create, validate, review, and maintain Claude Code plugins with repeatable standards.
- Daily operator tools — plugins such as
gws,safe-fetch,
nano-banana-pro, handoff, quick-notes, and session-retro cover common productivity, media, safety, continuity, and reflection workflows.
At the center is the fakoli trinity — three plugins designed to compose, each useful standalone:
| Plugin | Role | What It Does | |--------|------|--------------| | fakoli-flow | Workflow orchestration | Intent-driven pipeline: brainstorm → plan → execute → verify → finish. Wave-based dispatch with mandatory critic gates between every code-writing phase | | fakoli-crew | Specialist agents | 9 polyglot agents (TypeScript / Python / Rust) — architect, reviewer, security auditor, researcher, plugin engineer, integration specialist, documenter, infrastructure engineer, QA | | fakoli-state | Canonical project state | Local-first SQLite state engine for humans and AI coding agents. Lockable, evidence-backed work packets; event-sourced log with rebuildable SQLite projection; CLI + MCP surfaces, bidirectional GitHub Issues sync |
fakoli-flow defines how work moves. fakoli-crew defines who does the work. fakoli-state defines what is true. Together they form the canonical fakoli stack: orchestration + specialists + durable state.
A separate companion plugin, systems-thinking, runs multi-agent infrastructure analysis (discovery → extraction → synthesis) for decisions that affect the whole system — useful before the trinity starts work, but not part of the core stack.
Install any combination — each works standalone. The full marketplace gives you a path from idea to execution: evaluate the system context, plan work, assign it to specialists, capture durable handoffs, run safe tool-enabled workflows, and verify plugin changes through a shared repository health gate.
For prior-art context — what fakoli invented, what it reinvented, and where the moat is — see docs/POSITIONING.md.
---
Available Plugins
Google Workspace & Productivity
| Plugin | Description | |--------|-------------| | gws | Full Google Workspace via the gws CLI — 100 skills, 15 commands, 11 role-based agents, and 44 recipes spanning Gmail, Calendar, Drive, Docs, Sheets, Slides, Chat, and more. The most comprehensive Workspace plugin available for any AI assistant. | | notebooklm-enhanced | Programmatic control of Google NotebookLM — create notebooks, ingest PDFs and YouTube videos, generate podcasts and slide decks, and run end-to-end research workflows with a single command. |
AI & Media Generation
| Plugin | Description | |--------|-------------| | nano-banana-pro | Generate, edit, and remix production-ready images with Google Gemini 3 Pro. Includes a 5-agent PaperBanana pipeline (Retriever → Planner → Stylist → Visualizer → Critic) that iteratively refines images until they pass a quality threshold. | | fakoli-speak | Multi-provider TTS for Claude Code — stream any response as speech via /speak using OpenAI ($0.015/1K), Deepgram, ElevenLabs, Google Gemini (free), or macOS Say (free). Switch with /provider, track spending with /cost, toggle auto-narration with /autospeak. | | excalidraw-diagram | Generate .excalidraw files from natural language or by analyzing your codebase. Supports flowcharts, architecture diagrams, ER diagrams, and dependency graphs across four color themes — zero dependencies beyond Node.js 18. |
Security & Web
| Plugin | Description | |--------|-------------| | safe-fetch | Drop-in replacement for Claude's built-in WebFetch and WebSearch that runs content through a 6-layer sanitization pipeline before it touches the LLM. Neutralizes CSS-hidden text, zero-width Unicode, fake LLM delimiters, base64 payloads, and markdown exfiltration vectors. Security-team approvable. |
Development & Workflow
| Plugin | Description | |--------|-------------| | anvil-pulse | Live operator dashboard for long autonomous anvil runs — a dependency-free local web page showing active claims (actor, phase, elapsed, live lease countdown), the event feed, and per-claim stuck-state detection (healthy / quiet / possibly-wedged / lease-expired), plus an optional Claude Code statusline segment. Answers "is this still going, or is it wedged?" on any harness. | | cli-to-plugin | Convert any CLI with --help support into a Claude Code plugin: one skill per command group plus optional LLM-proposed workflow meta-skills. | | gate-router | Deterministic changed-path gate routing — map the files you changed to the verify commands this repo requires before shipping (docs -> strict build, shell -> bash -n, src -> suite), from a per-project rules file (.claude/gate-router.local.md) instead of session memory. | | recall-mode-verifier | Spec-independent breakage-probe skill — attack a change along fail-closed, malformed-input, resource-exhaustion, and state-drift axes (not its own tests). Verify-left pass for anvil execute and pre-PR review; reports, doesn't fix. | | windows-cli-hygiene | Advisory scanner for Windows/cross-platform CLI hazards — non-ASCII in printed strings (cp1252 crashes), hardcoded python3, heredoc backslash mangling, Node .cmd/.bat spawns, set -e in hooks. The scanner form of ship-loop's Windows discipline; wire it as a gate-router gate. | | fakoli-crew | Summon nine expert agent archetypes — polyglot architect, Staff-Engineer code reviewer, security auditor, API researcher, plugin engineer, integration specialist, documentation writer, infrastructure keeper, and QA validator — that work independently or as coordinated crews using wave-based orchestration with hook-enforced review gates. | | fakoli-flow | Intent-driven workflow orchestration — brainstorm, plan, and execute complex projects through coordinated specialist agents with a five-stage pipeline (brainstorm → plan → execute → verify → finish), critic gates, and evidence-based verification. Works best alongside fakoli-crew. | | fakoli-state | Local-first, runtime-neutral project state engine for humans and AI coding agents. Turn PRDs into lockable, evidence-backed work packets; coordinate multiple agents without conflicts. Event-sourced JSONL source of truth with a rebuildable SQLite projection, lease-based claims, evidence-gated completion, score-driven task expansion, and bidirectional GitHub Issues sync. | | systems-thinking | Analyze infrastructure decisions through discovery, extraction, and synthesis so hidden costs, dependencies, caveats, and risks are visible before implementation. | | marketplace-manager | Create and manage plugins without leaving Claude Code — scaffold new plugins from template with /add-plugin, validate manifests, regenerate registry indices, and install GitHub Actions workflows. The tool that maintains this marketplace. | | fakoli-style | Governed ledger of the Fakoli operating-model principles — tracks proven, asserted, and aspirational lifecycle statuses and generates a Markdown report from a single canonical source. | | fakoli-plugin-critic | Five specialist critic agents for Claude Code plugin development — agent-critic (frontmatter/color/tools), skill-critic (lazy-loading/no-fuzzy-detection), hook-critic (contract-aware/${CLAUDE_PLUGIN_ROOT}), mcp-critic (schema/actor-validation), structure-critic (manifest/CHANGELOG/version-sync). Audit any plugin with MUST FIX / SHOULD FIX / CONSIDER / NIT verdicts. | | ship-task | The mechanical tail of a one-PR-per-task loop, as a single command. After review passes, /ship pushes the branch, opens a PR, waits for CI, squash-merges and deletes the branch, syncs the base, and optionally runs a post-merge step (e.g. anvil apply) — keeping the repetitive git/gh/CI/merge dance out of the context window. Makes no review decisions; stops if CI fails. | | ship-loop | The full shipping procedure as one invocable skill — sync + worktree isolation, ground-truth scoping, reality-faithful tests, an eight-angle adversarial review as the merge gate, squash-merge, and follow-up closure (promotions ledger, out-of-diff issues). Composes with ship-task for the mechanical tail. |
Continuity & Personal Workflow
| Plugin | Description | |--------|-------------| | handoff | Store durable cross-session project handoff notes shared across same-remote checkouts. | | quick-notes | Capture, search, edit, and export dictation-friendly personal notes. | | session-evals | Mine coding-agent sessions (Claude Code, Codex, OpenClaw, Cursor CLI) into local-model eval suites executable via anvil-serving — retro-first mining, deterministic checks, routing evidence. | | session-retro | Analyze local Claude Code and Codex session logs into actionable retrospectives. |
---
Quick Start Examples
# Search your Gmail inbox
/triage
# Generate a hero banner for your README
/generate-image "Hero banner with bold headline 'Ship Faster' on dark gradient" --aspect 16:9 --size 2K
# Fetch a webpage without prompt-injection risk
/fetch https://docs.anthropic.com/en/docs/about-claude/models/overview
# Create an architecture diagram from your codebase
/excalidraw Diagram the architecture of this project
# Read the last Claude response aloud (defaults to OpenAI TTS)
/speak
# Switch TTS provider
/provider deepgram
# Scaffold a new plugin
/add-plugin my-new-plugin
# Turn a PRD into claimable, evidence-backed work packets
fakoli-state init --name "My Project"
# (author .fakoli-state/prd.md by hand — see plugins/fakoli-state/docs/prd-template.md)
fakoli-state prd parse # PRD → requirements + tasks
fakoli-state prd review # gate: draft → reviewed
fakoli-state prd review --approve # gate: reviewed → approved
fakoli-state plan && fakoli-state score && fakoli-state review tasks # generate + score + ready
fakoli-state claim T001 # lockable claim with lease + heartbeat
fakoli-state submit T001 \
--commands "pytest" --files-changed src/foo.py # record evidence (gate input)
fakoli-state apply T001 --approve # promote to done with audit trail
---
For Plugin Authors
Create Your First Plugin in 5 Steps
1. Scaffold from template
cp -r templates/basic plugins/your-plugin-name
2. Fill in the manifest — edit .claude-plugin/plugin.json with your plugin's name, version, description, and declared capabilities.
3. Build your capabilities — add skills in skills/, slash commands in commands/, agents in agents/, or hooks in hooks/.
4. Run the repository health gate before pushing
./scripts/check-all.sh
This is the same combined command CI runs. It performs marketplace validation, path-resolution and hook-safety checks, affected plugin tests, and the hook validation suite. A zero exit means every step passed; any failed step prints the failing layer and exits non-zero, which also fails CI.
5. Submit a pull request — the CI pipeline will validate your plugin automatically. See the Contributing Guide for review criteria.
Plugin Structure
your-plugin/
├── .claude-plugin/
│ └── plugin.json # Plugin manifest (required)
├── skills/ # Skill context files
│ └── skill-name/
│ └── SKILL.md
├── commands/ # Slash command definitions
│ └── command-name.md
├── agents/ # Sub-agent configurations
│ └── agent-name.md
├── hooks/ # PreToolUse / PostToolUse hooks
│ └── hook-name.md
├── scripts/ # Supporting scripts (bash, python, node)
├── README.md # Plugin documentation (required)
└── LICENSE # License file
Validation Pipeline
Every pull request runs the same combined repository health gate contributors should run locally:
./scripts/check-all.sh
The command runs scripts/validate.sh, scripts/test-path-resolution.sh, affected plugin tests, and the hook validation suite in sequence. It exits 0 only after all checks pass. A non-zero exit stops at the failing layer and fails the CI job.
The CI workflows also run repository-specific follow-up checks:
| Check | What it validates | |-------|-------------------| | validate.yml | Combined health gate, registry drift, and handoff path resolver | | pr-check.yml | Combined health gate, registry drift, handoff path resolver, and pull request registry preview | | update-index.yml | Auto-regenerates registry/index.json on merge to main |
The schema lives in schemas/plugin.schema.json. For quick manifest-only iteration, run ./scripts/validate.sh plugins/your-plugin-name; before pushing, run ./scripts/check-all.sh.
---
Documentation
| Guide | Description | |-------|-------------| | Contributing Guide | How to safely contribute plugins — review process, security requirements, and merge criteria | | Create Your Own Marketplace | Fork this repo and run your own private or public plugin marketplace | | Plugin Guidelines | Best practices for plugin structure, skill authoring, and command design | | Testing Standards | Requirements for plugin test coverage | | Anthropic Plugin Docs | Official Claude Code plugin documentation from Anthropic |
---
Repository Structure
fakoli-plugins/
├── .claude-plugin/ # Marketplace-level manifest
├── .github/workflows/ # CI: validate, update-index, pr-check, schema-drift
├── plugins/ # All active plugins (24)
├── archive/ # Archived / deprecated plugins
├── registry/ # Auto-generated plugin index (do not edit manually)
├── schemas/ # JSON Schema definitions for manifests
├── scripts/ # validate.sh, generate-index.sh, and other tools
├── templates/ # Starter templates for new plugins
│ └── basic/ # Standard plugin scaffold
├── assets/ # Marketplace assets (banner, logos)
└── docs/ # Guides and documentation
---
Archived Plugins
Some plugins have been archived and are no longer actively maintained. They remain available in the archive/ directory for reference.
---
License
This marketplace is licensed under the MIT License. Individual plugins may have their own licenses — check each plugin's LICENSE file.
---
<p align="center">Built and maintained by <a href="https://github.com/fakoli">@fakoli</a></p>








