Skip to content

Latest commit

 

History

History
102 lines (74 loc) · 3.76 KB

File metadata and controls

102 lines (74 loc) · 3.76 KB

Claude Code / Factory CLI Setup

Personal configuration for Claude Code and Factory CLI, including custom commands, skills, and coding guidelines.

Folder structure

Key folders

  • commands/: Custom slash commands for the CLI.
  • droids/: Custom droids (agents) configuration.
  • hooks/: Automation hooks that run on certain events.
  • rules/: Prescriptive rules the agent should follow.
  • skills/: Reusable skills the agent can load.

Key files

  • .gitignore: Git ignore rules for this repo.
  • AGENTS.md: High-level agent instructions.
  • AGENTS-verbose.md: Expanded/verbose agent guidance.
  • memories.md: Non-prescriptive preferences/decisions.

AI Coding Setup — MCP Servers

MCP servers that are useful to install when setting up an AI-augmented coding environment / CLI (e.g., Claude Code / Codex / Factory).

MCP servers recommended to install

  • morph-mcp — local code editing + semantic repo discovery (edit_file, warpgrep_codebase_search)

  • context7 — updated library/framework documentation retrieval

    • Avoids relying on outdated, generic library APIs.
    • MCP Repo: https://github.com/upstash/context7
    • Endpoint: https://mcp.context7.com/mcp
    • Needs: Context7 API key/token (set via headers in your MCP config)
  • nano-banana-pro — diagram/image generation (Gemini AI - Nano Banana Pro Image model)

  • perplexity — web research / reasoning

    • Run: npx -y perplexity-mcp
    • Needs: PERPLEXITY_API_KEY
  • brave-search — web search

    • Run: npx -y @brave/brave-search-mcp-server
    • Needs: BRAVE_API_KEY
  • chrome-devtools — browser-based testing / debugging

    • Run: npx -y chrome-devtools-mcp@latest
  • grep-mcp — public GitHub/code search

    • Run: uvx grep-mcp

AGENTS.md - Coding Guidelines

Commands

/commit-files

Commits staged files using Conventional Commits format:

  • Reviews staged changes with git diff --cached
  • Generates a structured commit message: <type>[scope]: <description>
  • Supports types: feat, fix, chore, docs, refactor, perf, test, etc.

/improve-prompt

Takes a file path and improves the AI prompt contained within it.

Usage: /improve-prompt <file-name>

Skills

mgrep

Semantic grep for local files - superior to built-in search tools like grep.

mgrep "What code parsers are available?"           # Search current directory
mgrep "How are chunks defined?" src/models         # Search specific directory
mgrep -m 10 "query"                                # Limit results

Installation: https://github.com/mixedbread-ai/mgrep

Features:

  • Natural-language semantic search
  • Multilingual & multimodal (code, images, PDFs)
  • Web search with --web flag
  • Background indexing via mgrep watch
  • Reduces agent token usage by 2x

Other useful skills

  • prompt-refiner — rewrites rough task descriptions into clearer, more actionable prompts.
  • prompt-refiner-gpt — prompt refinement patterns tuned for GPT/Codex.
  • react-useeffectuseEffect best practices from React docs (and when not to use Effects).
  • skill-creator — guidance for creating/updating skills (structure, metadata, resources).
  • web-design-guidelines — UI/UX + accessibility review against Web Interface Guidelines.

Optional skills to add