claude-baseline: Instructions file for Claude Code

CLAUDE.md

This file configures Claude Code behavior for this project. It is also a template: copy it into any new project, replace the placeholders, and delete sections that do not apply to your stack.

---

Project

• Owner: [YOUR NAME OR TEAM]
• Stack: [LANGUAGE / FRAMEWORK / DATABASE]
• Package manager: [npm / yarn / pnpm / pip / cargo / go]
• Test runner: [vitest / jest / pytest / go test / cargo test]
• Formatter: enforced automatically by .claude/hooks/format.sh

---

Code Style

Keep the block that matches your stack and delete the rest.

TypeScript / JavaScript

• Strict TypeScript; explicit return types on all exported functions
• const over let; never var
• async/await only; no callbacks, no raw promise chains
• Named exports only; no default exports
• Template literals over string concatenation

Python

• Type hints on all function signatures
• PEP 8 compliance (enforced by Ruff or Black)
• f-strings over .format() or % formatting

Go

• Standard gofmt style (enforced automatically)
• Errors returned and handled explicitly; never ignored with _
• Interfaces defined at the point of use, not the point of implementation

Bash

• set -euo pipefail on every script
• local for all function-scoped variables
• Quote all variable expansions: "$VAR", not $VAR
• Use [[ not [ for conditionals
• Snake_case function names

---

Naming

• TS/JS: camelCase vars/functions, PascalCase classes/types/components
• Python: snake_case vars/functions, PascalCase classes
• Booleans: isActive, hasPermission, canRetry
• Functions: verb-noun pairs: fetchUser(), parseToken(), validateInput()
• No single-letter variables except loop counters (i, j, k)
• Banned names: data, info, tmp, flag, val, obj, result
• Names are self-documenting. If a name needs a comment to explain it, rename it.

---

Engineering Constraints

• One function, one responsibility. Max 30 lines of logic. Max 3 parameters; use an options object or dataclass beyond that.
• Max 2 levels of nesting; use early returns.
• Never silently swallow exceptions. Log with: timestamp, error type, stack trace, and context.
• Validate and sanitize all external input. Assume hostile data at every system boundary (user input, API responses, file reads, environment variables).
• Never hardcode secrets, keys, or tokens. Environment variables only.
• Flag any injection risk: SQL, command, template literal, path traversal.
• Least privilege on all permission and scope decisions.