Write a short spec (Objective, Acceptance Criteria, edge cases, non-goals) before touching code. Use when asked to build a feature and the requirements are fuzzy or underspecified.

Use when reducing codebase complexity — find dead code with vulture/ruff, collapse single-implementation interfaces, inline premature indirection. The 'could the next reader delete this?' test.

Write a handoff note (where things stand, decisions made, exact next step) so a task can resume cleanly across sessions or agents. Use at the end of any work session, or when resuming one.

Before reporting any task done, actually run the code, read the diff, and confirm the output — never claim completion from memory. Use at the end of every task.

Review your own diff line-by-line before calling work done — read the actual diff, run the checks, and fix your own issues first. Use before reporting any change complete.

Respond to a leaked secret — rotate immediately, purge from git history, and prevent recurrence. Use the moment a secret is found committed or exposed.

Do exactly what was asked — no gold-plating, no unrequested refactors or dependency additions — and ask before expanding scope. Use to keep changes small, reviewable, and reversible.

Use when changing code structure without changing behavior — the mechanical refactoring catalog (extract, rename, inline, move), small-step discipline, tests stay green throughout.

Classify review findings into blocking vs non-blocking (nit) with clear criteria so reviews stay fast and fair. Use when writing or responding to review feedback.

Review LLM/agent-generated code with extra skepticism — verify claims, check for subtle bugs and invented APIs, and confirm it actually runs. Use for any AI-authored diff before trusting it.