Skip to content

context.md

Latest commit

 

History

History
51 lines (38 loc) · 2.37 KB

context.md

File metadata and controls

51 lines (38 loc) · 2.37 KB

Global AI Context

Single source of truth for all AI coding assistants. Managed by: ~/.config/ai/scripts/ai-config-sync.sh

Machine

  • MacBook Pro M3 Pro, 36 GB RAM, 512 GB SSD, macOS Sequoia
  • Shell: zsh

AI Tools Installed

  • IDEs: Cursor, Visual Studio Code, Windsurf
  • IDE Plugins: Cline, Claude Code, GitHub Copilot
  • Browser: Arc

Runtime Management

  • Python and Node.js versions are managed per project using mise.
  • Never use system Python/Node. Check for .mise.toml in the project root first.
  • If absent, run mise use python@<version> or mise use node@<version> before installing dependencies.
  • If mise exec is not picking up site-packages outside a project directory, invoke Python via its full path: ~/.local/share/mise/installs/python/<version>/bin/python.

Coding Preferences

Languages & Linting

  • Python: format with black, lint with flake8, scan with bandit
  • TypeScript/JavaScript: lint with eslint
  • Shell: target POSIX-compatible bash; use shellcheck conventions (quote variables, use [[ ]], avoid eval)
  • IaC: Terraform with consistent naming (snake_case resources, kebab-case tags)

Architecture

  • Prioritize cloud-native, containerized code with Docker/K8s compatibility
  • Prefer 12-factor app patterns (env-based config, stateless processes, disposable containers)
  • Use 1Password CLI or pass for secrets — never hardcode or commit credentials

Design Axiom — Balance

  • Every addition must justify its cost across four dimensions: completeness (does it fill a real gap?), security (does it reduce attack surface?), convenience (does it save meaningful time?), and resource cost (memory, CPU, disk, network)
  • When dimensions conflict, prefer: security > resource cost > convenience > completeness
  • Periodically audit for bloat — remove anything that no longer earns its keep

Quality

  • Validate at system boundaries (user input, external APIs) — trust internal code
  • Write tests for business logic and edge cases, not for framework glue
  • Handle errors explicitly — no silent catches, always log or propagate

Style

  • Commit messages: imperative mood, concise subject (<70 chars), explain "why" not "what"
  • PR descriptions: summary bullets + test plan
  • Never add Co-Authored-By trailers to commit messages
  • Never reference AI tools, AI assistance, or AI-generated content in committed code or commit messages