Freelancer Toolkit
TSiez/freelancer-toolkit
Automação de workflows freelance: onboarding, conteúdo, análise competitiva e relatórios
Ver no GitHubKit completo de automação para freelancers em Claude Code, incluindo agents especializados, comandos de linha (/), skills e hooks para onboarding de clientes, geração de conteúdo, análise competitiva e gestão de projetos. Integra-se com GitHub via MCP para consolidar todo o fluxo de trabalho solo dentro de um único ambiente.
Instalação
npx convez-ai TSiez/freelancer-toolkit1 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
Casos de uso
- Onboarding automatizado de novos clientes com geração de briefs e estrutura de pastas
- Geração e revisão de conteúdo com análise de SEO e qualidade
- Análise competitiva estruturada e comparativa com seu produto
- Geração de relatórios semanais profissionais para clientes
- Gerenciamento integrado de repositórios GitHub dentro do fluxo de trabalho
Veredito

Toolkit robusto e bem arquitetado para o freelancer que usa Claude Code. Combina agents, hooks e comandos em um pacote coerente; a integração MCP com GitHub agrega valor real. Requer familiaridade com Claude Code, mas entrega ROI alto em economia de tempo em tarefas repetitivas de gestão e conteúdo.
README
A tradução para o português deste README ainda está na fila.
Freelancer Toolkit
A freelancer's toolkit for client onboarding, content creation, and project management — packaged as a Claude Code plugin. It bundles slash commands, specialized research/review agents, a client-brief skill, content-quality hooks, and a GitHub MCP integration so you can run your whole solo-freelance workflow from inside Claude Code.
What's inside
| Type | Name | Purpose |
|---|---|---|
| Skill | client-brief-generator | Generate a comprehensive client brief and save it to client-briefs/ |
| Command | /onboard-client | Full new-client onboarding: brief + project folders + next steps |
| Command | /weekly-report | Generate a professional weekly status report into reports/ |
| Command | /research-and-write | Research a topic, write an 800–1200-word blog post, then review it |
| Command | /competitor-analysis | Scout a competitor and compare them to your product |
| Agent | research-agent | Finds 3–5 credible sources and returns structured research notes |
| Agent | review-agent | Reviews a draft for SEO, accuracy, and quality with a 1–10 score |
| Agent | competitor-scout-agent | Profiles a competitor: pricing, features, positioning, feedback |
| Agent | competitor-compare-agent | Compares a competitor to your product and finds gaps to pursue |
| Hook | quality-check | PostToolUse content check — warns on thin content and placeholder text |
| Hook | session-summary | Stop hook — prints a summary of files changed and commands run each session |
| MCP | github | Manage repos, issues, and PRs for project work |
Installation
This repository is a single-plugin marketplace, so you install it by adding the marketplace and then installing the plugin from it.
From a local clone
git clone https://github.com/sieztriciaann/my-freelancer-plugin.git
In Claude Code:
/plugin marketplace add ./my-freelancer-plugin
/plugin install freelancer-toolkit@freelancer-toolkit-marketplace
From GitHub directly
/plugin marketplace add sieztriciaann/my-freelancer-plugin
/plugin install freelancer-toolkit@freelancer-toolkit-marketplace
Then restart Claude Code (or run /reload-plugins) so the commands, agents, hooks, and MCP servers register.
Verify it loaded:
/help # the freelancer-toolkit commands should appear
/mcp # the github server should be listed
The MCP servers won't connect until you set their environment variables — see How MCP works below.
Slash commands
All commands are namespaced under the plugin; type / in Claude Code and they'll autocomplete. Each prompts you for anything it needs, so you can run them with no arguments.
/onboard-client
Walks through onboarding a new client end to end.
- You provide: client name and services (it asks for the rest of the brief fields).
- What it does: generates a full client brief, scaffolds the project folders (
client-briefs/,content/drafts/,content/published/,reports/), saves the brief toclient-briefs/<client-name>-brief.md, and summarizes next steps. - Example:
/onboard-client→ answer the prompts → folders + brief created.
/weekly-report
Generates a professional weekly status report for a client.
- You provide: work completed, content created, estimated hours, blockers, and next week's plan (it infers from
content/and prior reports where possible). - What it does: produces a clean, sectioned report and saves it to
reports/weekly-report-<YYYY-MM-DD>.md. - Example:
/weekly-report→ fill in the week's activity → report saved with today's date.
/research-and-write
Produces a researched, reviewed blog post.
- You provide: a topic (optionally target audience and angle/keyword).
- What it does: runs the
research-agentto gather 3–5 credible sources, writes an 800–1200-word post grounded in them, then runs thereview-agentto score it (SEO / accuracy / quality, 1–10) and applies the priority fixes. Offers to save the result intocontent/drafts/. - Example:
/research-and-write→ "How freelancers should price retainers" → drafted, reviewed, and scored post with sources.
/competitor-analysis
Analyzes one competitor against your product.
- You provide: the competitor's name/URL and a short description of your product.
- What it does: runs the
competitor-scout-agentto profile the competitor (pricing, features, positioning, customer feedback), then thecompetitor-compare-agentto compare it to your product, and outputs a tight three-section summary — What they do well / Where I have an edge / Opportunities to pursue. Offers to save intoreports/. - Example:
/competitor-analysis→ "Bonsai, hellobonsai.com" + your product blurb → comparison report.
The client-brief-generator skill
The brief generator is also available as a standalone skill. It triggers automatically when you say things like "create a client brief" or "onboard a new client," or you can invoke it directly. It collects ten fields (client name, industry, services, target audience, brand voice, platforms, budget, timeline, competitors, success metrics), outputs a markdown table, and saves it to client-briefs/<client-name>-brief.md. /onboard-client uses this skill under the hood.
How MCP works in this plugin
MCP (Model Context Protocol) servers give Claude live tools to act on external services. This plugin ships a .mcp.json at its root declaring one HTTP MCP server — github — under the standard mcpServers key. Claude Code auto-discovers this file when the plugin loads.
No secrets are stored in the plugin. The server reads its credential from an environment variable (${GITHUB_TOKEN}) that Claude Code expands at runtime. You set the variable once on your machine; the plugin only references it by name. After setting it, restart Claude Code and run /mcp to confirm the server connects.
GitHub MCP setup
Connects to the official remote GitHub MCP server (https://api.githubcopilot.com/mcp/) to manage repos, issues, and pull requests.
1. Get a personal access token
- Sign in to github.com.
- Go to Settings → Developer settings → Personal access tokens → Fine-grained tokens (or use a classic token).
- Click Generate new token, name it (e.g.
claude-mcp), set an expiration, and select the repositories it can access. - Grant the scopes you need — typically Contents, Issues, and Pull requests (read/write). For classic tokens, the
reposcope covers these. - Click Generate token and copy it immediately — GitHub only shows it once.
Treat this token like a password. Scope it to only the repos and permissions you need.
2. Set the environment variable
macOS / Linux:
export GITHUB_TOKEN="ghp_your-token-here"
Then reload your shell. Windows (PowerShell):
setx GITHUB_TOKEN "ghp_your-token-here"
3. Verify — restart Claude Code and run /mcp to confirm the github server connects. If it fails, check the token hasn't expired and has access to the target repositories.
How hooks work in this plugin
Hooks are scripts Claude Code runs automatically on certain events. This plugin registers them in hooks/hooks.json, which Claude Code auto-discovers on load. The scripts are invoked via bash "${CLAUDE_PLUGIN_ROOT}/hooks/<script>.sh", where ${CLAUDE_PLUGIN_ROOT} resolves to wherever the plugin is installed — so they work regardless of install location.
Requirement: the hooks are Bash scripts. They run anywhere
bashis on the PATH (macOS, Linux, and Windows with Git Bash or WSL). On a pure-PowerShell Windows setup without bash, the hooks won't execute (everything else in the plugin still works). The scripts preferjqfor parsing but fall back togrep/sedwhenjqisn't installed, and they always exit0so they never block your work.
quality-check — PostToolUse
Runs after every Write or Edit. For content files (.md, .txt, .html) it:
- warns if a file under a
content/directory is thin (under 200 words), and - flags leftover placeholder text (
lorem ipsum,TODO) in any content file.
Warnings print to your session; they're advisory and never block the edit.
session-summary — Stop
Runs when Claude finishes responding. It reads the session transcript and prints a short summary of what was accomplished — the files created or edited (deduplicated) and the number of shell commands run.
Configuration reference
| Variable | Used by | Required for |
|---|---|---|
GITHUB_TOKEN | github MCP | Authenticating GitHub requests |
Project structure
my-freelancer-plugin/
├── .claude-plugin/
│ ├── plugin.json # plugin manifest
│ └── marketplace.json # single-plugin marketplace manifest
├── .mcp.json # GitHub MCP server config
├── skills/
│ └── client-brief-generator/SKILL.md # client-brief-generator skill
├── agents/
│ ├── research-agent.md # research-agent (sources + notes)
│ ├── review-agent.md # review-agent (SEO/accuracy/quality)
│ ├── competitor-scout-agent.md # competitor-scout-agent (profile)
│ └── competitor-compare-agent.md # competitor-compare-agent (comparison)
├── commands/
│ ├── onboard-client.md # /onboard-client
│ ├── weekly-report.md # /weekly-report
│ ├── research-and-write.md # /research-and-write
│ └── competitor-analysis.md # /competitor-analysis
└── hooks/
├── hooks.json # PostToolUse + Stop registration
├── quality-check.sh # content quality check (PostToolUse)
└── session-summary.sh # session summary (Stop)
Author
Tricia Siez · sieztriciaann@gmail.com