Skip to content
/

wyndomb/claude-code-starter

wyndomb-cc-starter · wyndomb/claude-code-starter · ★ 6 · last commit 2026-01-29

A productivity-workflow Claude Code starter for knowledge workers shipping daily planning, research, standup, and weekly synthesis commands with pre-wired Perplexity + Notion + Obsidian MCP.

Best whenClaude Code's highest-value use for most people is personal productivity workflows (planning, research, synthesis), not software development automation.
Skip ifOver-engineering from the start, Walls of text without structure
vs seeds
agent-os(minimal automation, markdown scaffold) but wyndomb's template is oriented toward personal productivity (daily planning,…
Primitive shape 12 total
Commands 7 Subagents 1 MCP tools 4
00

Summary

wyndomb/claude-code-starter — Summary

wyndomb/claude-code-starter is a productivity-workflow Claude Code starter oriented toward knowledge workers rather than software developers. It ships 7 slash commands (daily-planning, meeting-prep, project-status, research-deep-dive, standup, productivity-analyzer, weekly-synthesis) and 1 named subagent (weekly-synthesis) with a pre-configured Perplexity and Firecrawl MCP integration. The CLAUDE.md is a deliberately incomplete template requiring the user to fill in their role, priorities, and communication preferences. The central value proposition is time savings for recurring knowledge-work tasks: "Saves 15-30 min/day" (daily planning), "Saves 2-3 hours" (research), "Saves 15-20 min per switch" (project status). No hooks, no tests, no git automation — the framework is maximally simple with minimal behavioral enforcement.

differs_from_seeds: Closest to agent-os (markdown scaffold, zero automation) but oriented toward personal productivity rather than software engineering. Unlike agent-os's standards/ content for code quality, wyndomb's value is in the weekly-synthesis agent that reads daily plan files across a week and generates reflection. No seed in the reference set targets knowledge-worker productivity workflows. Compared to claude-conductor (error ledger as first-class artifact), wyndomb focuses on future planning artifacts rather than error tracking.

01

Overview

wyndomb/claude-code-starter — Overview

Origin

Personal Claude Code starter published by wyndomb. 6 GitHub stars, no license, language classification: null (markdown-only). Last pushed 2026-01-29. Single contributor.

Philosophy

The README positions this as a 15-minute setup for knowledge workers:

"Get started with Claude Code in 15 minutes. Clone, customize, use."

The CLAUDE.md template's preamble sets the philosophical intent:

"What This File Does: This is Claude's instruction manual for working with you. It persists across sessions, so you're not starting from zero every time. Think of it as teaching Claude your work style, preferences, and decision-making patterns."

"How to Use This Template: Start minimal. Use this as-is for your first week, then customize based on what Claude actually needs to know about how you work."

The template explicitly names anti-patterns to watch for (over-engineering, getting stuck on options) and asks users to fill in their own failure modes.

Target User

Knowledge workers (not necessarily developers): analysts, consultants, product managers, writers. The commands are named daily-planning, meeting-prep, standup, productivity-analyzer — clearly productivity-workflow focused. The research-deep-dive command specifically references Perplexity MCP integration.

Weekly Synthesis Agent

The only agent is weekly-synthesis, designed to run automatically on Fridays:

"Automatic Friday reflection compiling accomplishments, blockers, and next week's focus (saves 45 min/week)"

The agent reads daily-plans/ files, projects/ files, git commits, and file changes to compile a weekly synthesis.

MCP Pre-Configuration

The .mcp.json file is pre-configured with 4 MCP servers (Firecrawl, Perplexity, Notion, Obsidian), unusual for such a simple starter — reflecting the research-workflow orientation.

02

Architecture

wyndomb/claude-code-starter — Architecture

Distribution

Standalone GitHub repository. Manual copy of files to user's workspace.

Install Method

git clone https://github.com/wyndomb/claude-code-starter
# Or download as ZIP

# Copy CLAUDE.md to workspace
cp CLAUDE.md ~/your-workspace/

# Copy desired commands
mkdir -p ~/your-workspace/.claude/commands/
cp .claude/commands/daily-planning.md ~/your-workspace/.claude/commands/

Directory Tree

wyndomb/claude-code-starter/
├── .claude/
│   ├── agents/
│   │   └── weekly-synthesis.md    (Friday reflection agent)
│   └── commands/
│       ├── daily-planning.md
│       ├── meeting-prep.md
│       ├── productivity-analyzer.md
│       ├── project-status.md
│       ├── research-deep-dive.md
│       ├── standup.md
│       └── weekly-synthesis.md   (also available as command)
├── .gitignore
├── .mcp.json                      (Firecrawl, Perplexity, Notion, Obsidian)
├── CLAUDE.md                      (customizable template)
├── README.md
├── REPOSITORY-SUMMARY.md
├── docs/
│   ├── claude-md/                 (CLAUDE.md design docs)
│   ├── customization-guide.md
│   ├── daily-plan-template.md
│   ├── mcp-setup-guide.md
│   ├── project-template.md
│   └── troubleshooting.md
├── examples/                      (sample outputs)
└── templates/                     (template files for daily plans, projects)

Required Runtime

  • Claude Code (with Anthropic API key)
  • Optional: Perplexity API key (for research-deep-dive command)
  • Optional: Firecrawl API key (for web scraping)
  • Optional: Notion integration (for Notion MCP)
  • Optional: Obsidian vault path (for Obsidian MCP)

Target AI Tools

Claude Code primarily. MCP servers add Perplexity, Firecrawl, Notion, Obsidian access.

MCP Config (.mcp.json)

{
  "mcpServers": {
    "firecrawl-mcp": { "command": "npx", "args": ["-y", "firecrawl-mcp"] },
    "perplexity-ask": { "command": "npx", "args": ["-y", "server-perplexity-ask"] },
    "notion": { "url": "https://mcp.notion.com/mcp" },
    "obsidian": { "command": "npx", "args": ["@mauricio.wolff/mcp-obsidian@latest"] }
  }
}
03

Components

wyndomb/claude-code-starter — Components

Slash Commands (7)

  • daily-planning.md — generates focused daily plan from calendar + yesterday's notes; produces time-blocked schedule with high/medium/low priority tasks; claims "saves 15-30 min/day"
  • meeting-prep.md — prepares for upcoming meetings by summarizing context and generating questions
  • productivity-analyzer.md — analyzes productivity patterns from daily plans and project files
  • project-status.md — instant context switch between projects; reads project files, recent commits, current blockers; "saves 15-20 min per switch"
  • research-deep-dive.md — comprehensive research using Perplexity MCP; synthesizes sources into structured report; "saves 2-3 hours"
  • standup.md — generates daily standup notes from git commits and project activity
  • weekly-synthesis.md — can also be invoked as a command (same capability as the agent)

Agents (1)

  • weekly-synthesis.md — Weekly Synthesis Specialist; reads daily-plans/ directory, projects/ directory, git commits; compiles: HIGH-IMPACT WINS, SOLID PROGRESS, LEARNING & GROWTH, DEFERRED INTENTIONALLY, BLOCKED, DROPPED; identifies time/energy patterns, decision quality, workflow friction; sets next week's THEME + TOP 3 PRIORITIES; writes output as markdown file; model: sonnet; color: blue; trigger: Friday or "I'm wrapping up for the week"

Hooks

None.

Skills

None (commands/agents only).

MCP Servers (4, pre-configured)

  • firecrawl-mcp — web scraping for research workflows
  • server-perplexity-ask — Perplexity AI search for research-deep-dive
  • notion — Notion workspace integration (HTTP MCP)
  • obsidian — Obsidian vault integration

Templates

  • docs/daily-plan-template.md — daily plan output format
  • docs/project-template.md — project status format
  • examples/ — sample command outputs

CLAUDE.md Template Sections

The CLAUDE.md requires user to fill in:

  • Project Overview (role, goals, tools)
  • Voice & Tone preferences
  • Key Preferences (decision-making patterns, AI collaboration style)
  • Anti-Patterns to Avoid
  • Quick Reference (key context always loaded)
  • Session Rhythm (how to start/end sessions)
  • Growth Areas (skills being developed)
05

Prompts

wyndomb/claude-code-starter — Prompts

Verbatim Excerpt 1: daily-planning.md command

Generate my daily plan for today.

Context to analyze:
- Yesterday's work: [describe what you worked on yesterday, or point to notes file]
- Today's calendar: [list meetings and time blocks, or share calendar access]
- Current priorities: [list your top 3-5 priorities this week]
- Blockers: [anything preventing progress on key tasks]

Output format:
1. **Focus Theme for Today:** [One sentence - what's the main goal?]

2. **Prioritized Tasks:**
   - High Priority (must complete today):
     - [Task 1] - [estimated time]
     - [Task 2] - [estimated time]

   - Medium Priority (if time allows):
     - [Task 3] - [estimated time]

Prompting technique: Structured context-injection template. The command is not a behavioral instruction to Claude but a user-facing template — the user fills in [brackets] with their actual context before running. Output format specified with explicit priority levels.


Verbatim Excerpt 2: weekly-synthesis.md agent (YAML frontmatter)

---
name: weekly-synthesis
description: Automated weekly reflection that analyzes daily plans and project
files to generate weekly accomplishments, blockers, and next week's strategic
focus. Use this agent every Friday or when the user wants to review their
week's progress and plan ahead. Examples:

<example>
Context: End of work week
user: "Can you generate my weekly synthesis?"
assistant: "I'm going to use the Task tool to launch the weekly-synthesis
agent to analyze your week and create your synthesis."
<commentary>User explicitly requested weekly synthesis.</commentary>
</example>

<example>
Context: Friday afternoon
user: "I'm wrapping up for the week"
assistant: "Before you finish, let me use the weekly-synthesis agent to
compile your accomplishments and plan for next week."
<commentary>Proactive use at natural reflection point.</commentary>
</example>
model: sonnet
color: blue
---

You are a Weekly Synthesis Specialist who helps knowledge workers reflect
on their week, identify patterns, and strategically plan ahead. Your
expertise lies in transforming daily activities into meaningful insights
about productivity, decision-making, and goal alignment.

Your primary responsibilities:

1. **Analyze the Past Week**: Review all available inputs...
   - Read daily plan files from the past 7 days (if available in
     `daily-plans/` directory)
   - Check project status files (if available in `projects/` directory)
   - Review conversation history for context about ongoing work
   - Examine git commits and file changes if relevant
   - Look for patterns in time allocation, energy, and focus

2. **Compile Accomplishments**: Identify and categorize what got done:
   - HIGH-IMPACT WINS: Significant achievements that moved the needle
   - SOLID PROGRESS: Steady advancement on important work
   - LEARNING & GROWTH: Skills developed or insights gained
   - Be specific with evidence - cite actual deliverables, not just effort

Prompting technique: Persona establishment ("Weekly Synthesis Specialist"). Structured output taxonomy with named categories (HIGH-IMPACT WINS, SOLID PROGRESS, LEARNING & GROWTH). Evidence-first framing ("cite actual deliverables, not just effort"). Two <example> blocks with <commentary> showing both explicit and proactive trigger patterns.

09

Uniqueness

wyndomb/claude-code-starter — Uniqueness

differs_from_seeds

The most distinctive positioning in this batch: wyndomb is the only "claude-code-starter" that explicitly targets knowledge workers rather than software developers. Closest seed is agent-os (markdown scaffold, minimal automation) but agent-os focuses on software development standards. The weekly-synthesis agent that reads a week's worth of daily plan files and git commits to generate structured reflection is a unique workflow not present in any seed or reference framework. The pre-configured 4-MCP bundle (Firecrawl + Perplexity + Notion + Obsidian) is unusual for a simple starter — it positions the framework as a research-and-knowledge-management tool. Compared to claude-conductor (error ledger + P0/P1 task tracking), wyndomb focuses on forward-planning artifacts (daily plans, weekly synthesis) rather than error tracking.

Positioning

"Get started with Claude Code in 15 minutes" — explicitly time-bounded setup for non-developer users. The README's time-savings claims (15-30 min/day, 2-3 hours/research) are the primary value proposition, framing Claude Code as a personal productivity assistant rather than a coding tool.

Observable Failure Modes

  1. No license: No license file in the repo. Users cannot legally determine usage rights.
  2. User-maintained context files: Daily plan files and project files must be created and maintained by the user. If user doesn't create daily-plans/ directory, weekly-synthesis agent has nothing to read.
  3. Template completeness: CLAUDE.md has many [bracket] placeholders. An unfilled template provides low-quality context to Claude.
  4. MCP API key requirements: Perplexity, Firecrawl, and Notion require API keys. The research-deep-dive command fails without Perplexity.
  5. Low maintenance signal: 6 stars, 1 contributor, last commit 2026-01-29. Abandoned risk.

Category Note

The only "productivity-workflow" starter in this batch — not a coding assistant toolkit but a knowledge-work amplifier. Represents a different market segment for Claude Code than the other 9 frameworks analyzed here.

04

Workflow

wyndomb/claude-code-starter — Workflow

Setup Phases

Phase Action Artifact
1. Install Claude Code Via VSCode or Cursor extension Claude Code configured
2. Clone and copy CLAUDE.md Copy to workspace, fill in role/priorities Customized CLAUDE.md
3. Install commands Copy desired .md files to .claude/commands/ Available slash commands
4. Set up MCP (optional) Add API keys for Perplexity, Firecrawl, Notion Research capabilities
5. Daily use Run commands as needed Plans, research, standups

Daily Knowledge-Work Loop

Command Timing Time Claimed Saved
/daily-planning Morning 15-30 min/day
/standup Before standup 10-15 min/day
/project-status Context switching 15-20 min/switch
/research-deep-dive As needed 2-3 hours/research
/meeting-prep Before meetings 15-20 min/meeting
/weekly-synthesis Friday 45 min/week

Weekly Synthesis Workflow

  1. End of Friday: user says "I'm wrapping up for the week" (or similar)
  2. Agent auto-triggers (description match) or user runs /weekly-synthesis
  3. Agent reads daily-plans/ directory (past 7 days)
  4. Agent reads projects/ directory
  5. Agent reviews git commits and file changes
  6. Agent generates structured synthesis markdown
  7. Agent writes file to user-specified location

Approval Gates

None. No formal gates, no blocking operations.

Artifacts

Artifact Where Produced By
Daily plan User-specified file or conversation daily-planning command
Weekly synthesis Markdown file weekly-synthesis agent
Research report Conversation output research-deep-dive command
Standup notes Conversation output standup command
06

Memory Context

wyndomb/claude-code-starter — Memory & Context

Memory Type

File-based. No database, no vector store.

Primary Memory Mechanism

CLAUDE.md is the sole persistent state. The template asks users to fill in role, priorities, communication preferences, and anti-patterns. This context persists across sessions automatically via Claude Code's CLAUDE.md loading.

Secondary Memory (implicit)

The weekly-synthesis agent creates markdown files. If users store these in their workspace, they accumulate as a record of weekly reflections. The daily-planning command suggests creating daily-plans/ and projects/ directories which the weekly-synthesis agent reads.

State Files

  • CLAUDE.md — user-customized workspace context
  • daily-plans/YYYY-MM-DD.md (user-created, agent reads)
  • projects/*.md (user-created, agent reads)
  • Weekly synthesis output files (agent-written)

Memory Persistence

Project-scoped via CLAUDE.md. Session-accumulated via daily plan files.

Compaction Handling

None — no pre/post compact hooks. No resilience mechanism.

Cross-Session Handoff

Partial — CLAUDE.md provides role/preference persistence. Daily plan files provide activity history if user maintains them.

Notable: No Structured Memory System

Unlike centminmod-cc-setup (dual-memory with explicit sync) or ccmemory (Neo4j graph), wyndomb relies entirely on the user to create and maintain context files. There is no automated memory capture, no sync workflow, and no recovery mechanism.

07

Orchestration

wyndomb/claude-code-starter — Orchestration

Multi-Agent Support

Minimal — 1 named subagent (weekly-synthesis). No orchestration protocol.

Orchestration Pattern

None — single commands, single agent. No coordination.

Multi-Model Usage

No — no model routing. Agent is pinned to sonnet but no multi-model strategy.

Isolation Mechanism

None.

Execution Mode

One-shot per command. The weekly-synthesis agent is the closest to a complex workflow, but it's single-agent, single-run.

Hooks

None.

Consensus Mechanism

None.

Crash Recovery

None.

Autonomy Level

Very low — each command requires user invocation. The weekly-synthesis agent has proactive trigger descriptions ("when user says they're wrapping up for the week") but relies on Claude Code's natural language agent matching, not explicit hooks.

08

Ui Cli Surface

wyndomb/claude-code-starter — UI / CLI Surface

Dedicated CLI Binary

No.

Local Web Dashboard

No.

IDE Integration

VSCode + Claude Code extension mentioned in README as primary setup path. Also mentions Cursor extension.

MCP Integration

4 MCP servers pre-configured in .mcp.json:

  • Firecrawl (web scraping)
  • Perplexity (AI search)
  • Notion (workspace)
  • Obsidian (notes vault)

This is the most MCP-forward configuration among "simple" starters in the batch.

Observability

None — no audit log, no token tracking, no notification hooks.

Git Automation

None.

Docs

docs/ contains:

  • customization-guide.md
  • mcp-setup-guide.md
  • troubleshooting.md
  • daily-plan-template.md
  • project-template.md
  • claude-md/ (design documentation for CLAUDE.md template)

Examples

examples/ contains sample outputs showing what commands produce.

Related frameworks

same archetype · same primary tool · same memory type

BMAD-METHOD ★ 48k

Provides a full agile delivery lifecycle with named expert-persona AI collaborators that elicit the human's best thinking rather…

Agent OS ★ 4.6k

Extracts implicit codebase conventions into token-efficient markdown standards files and injects them selectively into AI agent…

Claude Conductor ★ 367

Gives Claude Code a persistent, cross-linked, auto-analyzed documentation system so it retains codebase context across sessions.

Spec-Driver (Greenfield Spec-Driven Development) ★ 25

Prevents spec rot in AI-assisted development by making implementation changes flow back into evergreen, authoritative specs via…

Anthropic Knowledge Work Plugins ★ 16k

Role-specialized plugin bundles with live MCP connectors that turn Claude into a domain expert for enterprise knowledge workers.

Codex Integration for Claude Code (skill-codex) ★ 1.3k

Single Claude Code skill that handles Codex CLI invocation correctly (stdin blocking, thinking token suppression, session resume)…