Coding Skill

shiraz-coding-style

A coding-agent skill for **Claude Code**, (also has partial support for **Codex**), that enforces:

• strict TDD by default
• clarify before coding
• minimal, surgical diffs
• strong verification and explicit residual risk
• simplicity-first, anti-overengineering engineering discipline

This skill is designed for implementation work: bug fixes, features, refactors, migrations, UI changes, and multi-step engineering tasks where correctness matters more than speed.

What it does

`shiraz-coding-style` gives a coding agent a default operating methodology:

1. Clarify expected behavior and hidden product choices.
2. Resolve ambiguity from the repo, tests, docs, and existing behavior.
3. Ask only blocking clarifying questions.
4. Write RED tests first.
5. Confirm the tests fail for the correct reason.
6. Implement the smallest possible change to reach GREEN.
7. Refactor only while staying GREEN.
8. Review the final diff for regressions, overreach, and unnecessary complexity.

It combines:

• TDD-first implementation discipline
• verification-centered development
• simplicity-first, anti-overengineering principles
• surgical change management
• explicit handling of edge cases and assumptions

When to use it

Use this skill for:

• bug fixes
• new feature implementation
• refactors
• API changes
• UI changes
• migrations
• test creation or repair
• implementation-oriented code review
• multi-step engineering tasks

This skill is especially useful when you want the agent to:

• avoid guessing
• avoid overengineering
• keep diffs small and auditable
• make behavior explicit before coding
• prove correctness instead of asserting it

When not to use it as the primary workflow

This should not be the main workflow for:

• pure code reading with no implementation
• architecture brainstorming with no concrete change yet
• incident triage with no code change yet
• one-off research tasks
• documenta