A Claude-managed project template with automated workflows, skills, and development guidelines
Status: 🚧 Template - Fill TODOs and customize for your project
This is a minimal, stack-agnostic project template designed for use with Claude Code. It includes:
- ✅ Automated Skills: Context-aware guidance that activates based on your code and prompts
- ✅ Development Hooks: Auto-tracking of edited files and intelligent skill suggestions
- ✅ Strategic Agents: AI agents for planning, code review, and documentation
- ✅ Dev-Docs Pattern: Persistent task documentation that survives context resets
- ✅ Best Practices: Generic guidelines for structure, testing, deployment, and security
This template is intentionally generic with TODO: placeholders throughout. You'll customize it once you choose your tech stack.
Want to see a complete example? Check out .claude-example/ for a fully customized version using:
- Python 3.12+ with uv (fast package manager)
- Jupyter notebooks for data analysis
- pytest, mypy, ruff for quality checks
- All TODOs filled in with Python-specific values
View the Python/Jupyter example →
This example demonstrates exactly how to transform this generic template into a production-ready project for your stack.
Language: TODO: (e.g., TypeScript, Python, Rust, Go)
Framework: TODO: (e.g., Node.js, FastAPI, Actix, Gin)
Database: TODO: (e.g., PostgreSQL, MongoDB, Redis)
Testing: TODO: (e.g., Jest, pytest, cargo test)
Build Tool: TODO: (e.g., npm, cargo, poetry)
Deployment: TODO: (e.g., Docker, AWS, Vercel, Kubernetes)
# This repository is already cloned
cd /home/user/claude-setup
# Install dependencies (once you have a stack)
TODO: <install-command>Critical TODOs to fill:
- Replace
TODO: <project-name>everywhere with your project name - Replace
TODO: <short-description>with a 1-2 sentence description - Fill in tech stack placeholders in this README
- Update
.claude/skills/project-guidelines/SKILL.mdwith your stack details - Refine
.claude/skills/skill-rules.jsonwith project-specific patterns - Configure
.claude/hooks/post-tool-use-tracker.shfor your build system (optional) - Update
src/with your first module (remove any placeholder files)
Search for TODOs:
# Find all TODOs in the project
grep -r "TODO:" .claude/ README.md src/ docs/💡 Tip: See .claude-example/ for a complete Python/Jupyter example showing all customizations filled in.
The hooks use TypeScript. Install dependencies:
cd .claude/hooks
npm install
cd ../..Create task documentation for your first feature:
# In Claude Code, run:
/dev-docs my-first-featureThis creates:
.claude/dev/active/my-first-feature/my-first-feature-plan.md.claude/dev/active/my-first-feature/my-first-feature-context.md.claude/dev/active/my-first-feature/my-first-feature-tasks.md
This template uses Claude Code's automation features:
skill-activation-prompt.sh (UserPromptSubmit):
- Runs every time you send a prompt
- Matches your message against skill rules
- Suggests relevant skills automatically
post-tool-use-tracker.sh (PostToolUse):
- Runs after Edit, Write, or MultiEdit tools
- Tracks which files you've edited
- Can trigger builds/tests (TODO: configure for your stack)
project-guidelines:
- Comprehensive development guidelines (~500 lines)
- Progressive disclosure with resource files
- Auto-activates when editing code or planning features
- Customize with your stack, conventions, and patterns
How Skills Work:
- Defined in
.claude/skills/project-guidelines/SKILL.md - Activation rules in
.claude/skills/skill-rules.json - Triggered by file paths, keywords, or intent patterns
Specialized AI agents for complex tasks:
strategic-plan-architect:
- Creates comprehensive implementation plans
- Breaks work into phases with acceptance criteria
- Outputs plan, context, and task files
code-architecture-reviewer:
- Reviews code for quality and best practices
- Questions design decisions
- Saves detailed review to task directory
documentation-architect:
- Creates and updates documentation
- Gathers context from code and existing docs
- Produces developer-focused guides
Slash commands for workflows:
/dev-docs <task-name>:
- Creates structured task documentation
- Generates plan, context, and checklist files
- Persists across context resets
/dev-docs-update:
- Updates task docs before compaction
- Captures session progress and decisions
TODO: <project-name>/
├── src/ # Source code (TODO: customize structure)
├── tests/ # Tests (unit, integration, e2e)
├── docs/ # Project documentation
├── .claude/ # Claude Code infrastructure
│ ├── hooks/ # Automation hooks
│ │ ├── skill-activation-prompt.sh
│ │ ├── skill-activation-prompt.ts
│ │ ├── post-tool-use-tracker.sh
│ │ ├── package.json
│ │ └── tsconfig.json
│ ├── skills/ # Development guidelines
│ │ ├── project-guidelines/
│ │ │ ├── SKILL.md # Main skill file (~500 lines)
│ │ │ └── resources/ # Progressive disclosure files
│ │ │ ├── structure.md
│ │ │ ├── workflows.md
│ │ │ ├── testing.md
│ │ │ ├── docs.md
│ │ │ ├── release.md
│ │ │ └── security.md
│ │ └── skill-rules.json
│ ├── agents/ # Strategic AI agents
│ │ ├── strategic-plan-architect.md
│ │ ├── code-architecture-reviewer.md
│ │ └── documentation-architect.md
│ ├── commands/ # Slash commands
│ │ ├── dev-docs.md
│ │ └── dev-docs-update.md
│ ├── dev/active/ # Active task documentation
│ └── settings.json # Hook configuration
├── .gitignore
└── README.md # This file
# Use the strategic-plan-architect agent
/dev-docs implement-authenticationClaude creates:
implement-authentication-plan.md: Detailed implementation planimplement-authentication-context.md: Key decisions and filesimplement-authentication-tasks.md: Checklist for tracking progress
As you edit files:
- Hooks track your changes
- Skills auto-activate with guidance
- Follow checklist in
<task>-tasks.md
Launch the code review agent:
Use the Task tool to launch code-architecture-reviewer
Agent produces: implement-authentication-code-review.md with findings.
Launch the documentation agent:
Use the Task tool to launch documentation-architect
Agent creates/updates relevant documentation.
/dev-docs-updateUpdates task docs with progress, decisions, and next steps.
Edit .claude/skills/skill-rules.json:
{
"skills": {
"project-guidelines": {
"fileTriggers": {
"pathPatterns": [
"src/**/*.TODO:<ext>", // Add your file extensions
"TODO:<your-dir>/**/*" // Add your directories
],
"contentPatterns": [
"TODO:<import-pattern>", // e.g., "from myframework"
"TODO:<class-pattern>" // e.g., "class.*Service"
]
},
"promptTriggers": {
"keywords": [
"TODO:<domain-keyword>", // e.g., "authentication"
"TODO:<tech-keyword>" // e.g., "database"
]
}
}
}
}Edit .claude/hooks/post-tool-use-tracker.sh:
Add detection for your stack:
# Example: Detect package.json and track npm builds
if [[ -f "package.json" ]]; then
echo "npm run build" >> "$cache_dir/commands.txt"
fiEdit .claude/skills/project-guidelines/SKILL.md:
- Replace all
TODO:placeholders with actual values - Add language-specific examples
- Document your conventions and patterns
- Add common pitfalls to avoid
Create new skills in .claude/skills/<skill-name>/:
SKILL.md: Main guidance (~500 lines)resources/: Progressive disclosure files- Update
skill-rules.jsonto activate the skill
- Main
SKILL.md: ~400-500 lines - Resource files: ~100-300 lines each
- Use progressive disclosure (link to resources)
- Run
/dev-docs <task>for every non-trivial feature - Run
/dev-docs-updatebefore context compaction - Keep task docs up to date as you work
- Use
strategic-plan-architectfor planning - Use
code-architecture-reviewerafter implementation - Use
documentation-architectfor docs
To preserve hook tracking and docs alignment:
- Make all code changes through Claude Code
- If you edit manually, update task docs accordingly
/dev-docs-update- Review and update project guidelines
- Update dependency versions
- Refine skill rules based on actual usage
- Clean up completed tasks from
.claude/dev/active/
- Update architecture docs
- Revise testing strategy if needed
- Update deployment workflows
- Run code review agent on refactors
-
Check hook files are executable:
ls -la .claude/hooks/ chmod +x .claude/hooks/*.sh -
Verify
.claude/settings.jsonis valid JSON -
Check hook dependencies installed:
cd .claude/hooks && npm install
- Check
.claude/skills/skill-rules.jsonsyntax - Verify file paths match your project structure
- Test keyword triggers with your prompts
- Ensure
.claude/dev/active/directory exists - Check command exists in
.claude/commands/dev-docs.md - Verify you're using correct syntax:
/dev-docs <task-name>
- Claude Code Docs: docs.claude.com/claude-code
- Infrastructure Showcase: github.com/diet103/claude-code-infrastructure-showcase
- Repository: TODO:
- Issue Tracker: TODO:
- Documentation: TODO: (if applicable)
TODO: Add contributing guidelines if this is a team project
TODO: (e.g., MIT, Apache 2.0, Proprietary)
This template was created by transforming the Claude Code Infrastructure Showcase into a minimal, stack-agnostic starting point for any project using Claude Code.
Key Changes from Showcase:
- Removed all web/TypeScript-specific code
- Generalized skills to be language-agnostic
- Simplified hooks to two essential ones (no Stop hooks)
- Added TODO placeholders throughout for customization
- Created generic project-guidelines skill pattern
- Adapted agents to work across all stacks
Last Updated: 2025-11-15
Template Version: 1.0.0
Next Steps: Fill TODOs, choose your stack, and start building! 🚀