Claude Development Agents

December 31, 2025 ยท View on GitHub

Claude Code Projects

Claude Development Agents

A collection of Claude Code configurations for development workflows. Consolidates sub-agents, slash commands, and prompt snippets developed over the past few months.

Quick Reference

See INDEX.md for a comprehensive alphabetical listing of all 74 configurations with type badges and links.

Taxonomy

Agents (/agents/)

Full autonomous agent configurations for complex, multi-step tasks. Agents work independently within defined scopes and have YAML frontmatter with metadata.

Use when: The task requires autonomous investigation, multiple steps, or extended workflows.

Slash Commands (/slash-commands/)

Invokable commands that trigger specific workflows. Lighter than agents, designed for quick access to common operations.

Use when: You need to quickly invoke a focused workflow with /command-name.

Snippets (/snippets/)

Paste-able prompt fragments for steering conversations. Not invoked as commands - simply copied into prompts to redirect AI behavior.

Use when: You need to inject context or redirect an ongoing conversation without starting a formal workflow.

Orchestrators (/orchestrators/)

Meta-agents that coordinate multiple agents in sequence or parallel. Define workflows that invoke several agents in a single run.

Use when: You need a coordinated multi-agent workflow (e.g., code quality review team).

Crews (/crews/)

Curated groups of agents designed to work together on related tasks. Define team compositions with roles and interaction patterns.

Use when: Defining which agents collaborate on a domain (e.g., pre-release crew, documentation crew).

Directory Structure

agents/
  code-editing/        # Code manipulation, cleanup, debloating
  deployment/          # CI/CD and deployment automation
  development/         # Containerization, dev servers, stack alignment
  github/              # Backups, forks, licensing, collaboration
  ides/                # IDE integrations
  stylistic/           # Writing style enforcement
  subdevelopers/       # Single-bug debugging, feature additions
  web-design/          # Multi-agent creative web design system
  web-development/     # Responsive auditing, AI SEO, frontend optimization

slash-commands/
  codebase/            # Component search, dependency audit, consolidation
  development/         # Containerize, debug, devserver, backup, typescript-errors
  reviews/             # Stack, database, UI/UX, SEO, social, brand reviews
  workflow/            # Stuck handoffs, retrospectives, starting instructions

snippets/
  advisory/            # Smarter approach, right tool, efficiency checks
  complexity/          # Scope control, simplification requests
  debugging/           # Root cause focus, fire prevention
  session/             # Fresh starts, loop detection, corrections
  style/               # Emoji, docs, naming, structure preferences

orchestrators/         # Multi-agent workflow coordination (placeholder)

crews/                 # Curated agent team definitions (placeholder)

private/
  new-configs/         # Staging area for new configurations

Agent Categories

CategoryPurpose
code-editingComment removal, code cleanup
deploymentCI/CD setup, packaging, shipping
developmentContainerization, dev servers, stack advising
githubBackups, forks, licensing, collaboration
idesVS Code extensions and IDE tooling
stylisticWriting style enforcement (declaude)
subdevelopersSingle-bug debugging, feature additions
web-designMulti-agent creative web design system
web-developmentResponsive auditing, AI SEO, frontend optimization

Slash Commands

Development (/slash-commands/development/)

CommandPurposeCompanion Agent
/backupQuick repository backupgithub/backup-repo.md
/containerizeDocker setupdevelopment/containerize.md
/debugFocused bug fixingsubdevelopers/debugging.md
/devserverHot-reload setupdevelopment/devserver.md
/typescript-errorsResolve all TypeScript errors to zero-

Reviews (/slash-commands/reviews/)

Advisory commands that ask "how can we do this better?":

CommandPurpose
/brand-voiceReview copy against brand voice guidelines
/database-reviewDatabase optimization, indexes, relationships
/framework-leverageAre we fully using our framework's capabilities?
/seo-auditMeta tags, sitemap, robots.txt, structured data
/social-previewOpen Graph tags, Twitter cards, platform previews
/stack-reviewFresh eyes on technology stack choices
/ui-ux-reviewCSS efficiency and UX improvements

Codebase (/slash-commands/codebase/)

CommandPurpose
/component-searchFind existing UI components
/consolidateFind consolidation opportunities and inefficiencies
/dependency-auditAudit underutilized packages
/search-upgradeEvaluate search strategies (fuzzy, vector, full-text)
/template-checkFind existing templates

Workflow (/slash-commands/workflow/)

CommandPurpose
/retrospectiveReview for wheel reinvention
/starting-instructionNo wheel inventions reminder
/stuckGenerate handoff brief when stuck

Snippets

Session Control (/snippets/session/)

  • fresh-perspective - Request first-principles approach
  • previous-session-unhelpful - Context for retrying after failed session
  • this-was-the-issue - Correct a misdiagnosis
  • went-in-circles - Redirect when stuck in loops
  • wrong-direction - Signal need for pivot

Complexity Control (/snippets/complexity/)

  • stop-expanding-scope - Prevent scope creep
  • too-complex - Pull back from overcomplicated solutions

Debugging Focus (/snippets/debugging/)

  • dont-put-out-fires - Focus on fundamental problems
  • find-root-cause - Push past symptom-fixing

Style Directives (/snippets/style/)

  • balanced-structure - Prevent fragmented documents
  • match-brand-voice - Enforce specific brand voice
  • minimal-docs - Prevent over-documentation
  • no-all-caps - Enforce kebab-case naming
  • no-emojis - Prohibit emoji usage

Advisory Mode (/snippets/advisory/)

  • advisor-mode - Switch to advisory rather than executor mode
  • efficiency-check - Request efficiency analysis
  • hidden-relationships - Identify untapped data relationships
  • native-solution - Check if framework already solves this
  • right-tool - Question if we're using the right tool
  • smarter-approach - Ask "how can we do this smarter?"

Workflow

New configurations are drafted in private/new-configs/ and processed via the /sort command, which validates and moves them to appropriate categories.

Philosophy

This toolkit emphasizes:

  • No Wheel Inventions - Leverage existing solutions before building custom
  • Focused Scope - Each tool does one thing well
  • Fresh Starts - Recognize when to pivot rather than persist
  • Root Cause Thinking - Fix problems, not symptoms
  • Advisory Over Execution - Use AI as a strategic guide, not just a task executor

For more Claude Code projects, visit my Claude Code Projects Index.