TypeScript Strict Mode Migration Guide

What this prompt does

This prompt produces a phased plan to migrate a [project_size] [framework] project from [current_state] to strict TypeScript. The AI audits your tsconfig and recommends enabling [strict_flags] in order, categorises type errors by severity, fixes [priority_area] first, replaces [any_count]+ any types with generics and interfaces, adds type guards for runtime narrowing, centralises shared types in [types_location], configures eslint-typescript with [lint_rules], sets up path aliases, and adds a pre-commit hook blocking new any. It delivers a rollout the team follows over [timeline].

The structure works because it makes strictness incremental instead of a big-bang freeze. By enabling [strict_flags] in a sane order and killing any at the boundaries — [priority_area] — first, the riskiest runtime surprises get addressed early without halting feature work. The pre-commit guard is what stops the team quietly reintroducing any while the migration is in flight.

When to use it

• You have a loose or JavaScript codebase and want strict TypeScript without freezing development.
• You need a sane order for enabling [strict_flags] rather than flipping them all at once.
• You want to fix the highest-impact [priority_area] first.
• You need to systematically replace any types with real types.
• You want a pre-commit guard against new any additions.
• You are planning the migration across a fixed [timeline].

Example output

You get a phased plan: a tsconfig audit recommending [strict_flags] in order, an error inventory grouped by severity, a fix sequence starting with [priority_area], a strategy for replacing [any_count]+ any types, type guards, a [types_location] structure, eslint-typescript config with [lint_rules], path aliases, and a pre-commit hook — staged across [timeline].

Pro tips

• Enable [strict_flags] one at a time; strictNullChecks alone can surface a large error wave, so isolate it.
• Fix [priority_area] first — boundaries like the API layer are where any causes the worst runtime surprises.
• Be realistic about [any_count]; an estimate is fine, but expect the real number to shift as flags reveal more.
• Add the pre-commit any guard early so progress is not eroded while you work.
• Size the [timeline] to leave room for feature work; an aggressive schedule encourages any shortcuts.
• Centralise shared types in [types_location] so the team reuses definitions instead of duplicating them.