Extract a shared drift-review engine for brew/choco/skills review

[ChipWolf]

Context

Three drift-review scripts each re-implement the same engine: a tty guard (silent exit when non-interactive), a declared-vs-installed set diff, an interactive [a]dd / [r]emove / [i]gnore / [s]kip loop, and overlay-file appending. That's ~100+ duplicated lines per script; only the package parser and the overlay-append format genuinely differ.

This is a real seam with two-plus adapters: the interactive engine is one deep module, the per-tool parsing/appending are thin adapters over it.

Files

• home/dot_scripts/executable_brew-review (333 lines)
• home/dot_scripts/executable_choco-review (307 lines)
• home/dot_scripts/executable_skills-review (266 lines)

What to do

1. Load homebrew-management (.agents/skills/homebrew-management/SKILL.md) and update-skills (.agents/skills/update-skills/SKILL.md) first — brew-review and skills-review are governed by them.
2. Extract home/dot_scripts/review-lib (a sourced shell library) holding the shared engine: tty guard, declared-vs-installed diff, the 4-choice prompt loop, ignored-set filtering, and overlay append.
3. Reduce each *-review script to: source the lib + provide adapters (extract_declared, extract_installed, append_overlay) + invoke the engine.
4. Verify each script's interactive behaviour is unchanged (same prompts, same overlay edits) on a manual run.

Out of scope

• Adding new review tools.
• Changing the overlay file formats.