Review maintainability strictly (prompt)

curl -fsSL --create-dirs 'https://theskillvault.ai/api/skills/review-maintainability-strictly-prompt/raw' -o ~/'.claude/skills/review-maintainability-strictly-prompt/SKILL.md' && curl -fsSL --create-dirs 'https://theskillvault.ai/api/skills/review-maintainability-strictly-prompt/files/LICENSE.txt' -o ~/'.claude/skills/review-maintainability-strictly-prompt/LICENSE.txt'

curl -fsSL --create-dirs 'https://theskillvault.ai/api/skills/review-maintainability-strictly-prompt/raw' -o '.cursor/skills/review-maintainability-strictly-prompt/SKILL.md' && curl -fsSL --create-dirs 'https://theskillvault.ai/api/skills/review-maintainability-strictly-prompt/files/LICENSE.txt' -o '.cursor/skills/review-maintainability-strictly-prompt/LICENSE.txt'

curl -fsSL --create-dirs 'https://theskillvault.ai/api/skills/review-maintainability-strictly-prompt/raw' -o '.agents/skills/review-maintainability-strictly-prompt/SKILL.md' && curl -fsSL --create-dirs 'https://theskillvault.ai/api/skills/review-maintainability-strictly-prompt/files/LICENSE.txt' -o '.agents/skills/review-maintainability-strictly-prompt/LICENSE.txt'

{
  "mcpServers": {
    "theskillvault": {
      "url": "https://theskillvault.ai/mcp"
    }
  }
}

Code can pass every test and still leave the codebase worse. This prompt runs a demanding review of a branch that cares only about structure. It asks: is there a simpler shape that makes this whole change smaller?

When to use

• A feature works, but the diff feels tangled, long or full of special cases.
• AI agents have been adding code fast and you want a quality bar before merging.
• You want a second reviewer that will not wave things through because "it works".

When not to use

• You want bugs, security holes and edge cases found. Use review-code-prompt.
• You only want the fresh diff tidied in place. Use turn-messy-diff-into-clean-code-prompt.
• You want to redesign modules across the whole repo. Use review-codebase-architecture-prompt.

Steps

1. Commit your work, then open the branch in your AI coding tool.
2. Paste the prompt below. Change the base branch and the file-size limit if your project uses others.
3. Read the findings from the top. Structural ones matter more than naming.
4. Pick the restructurings you agree with and ask for them one at a time, with tests run after each.
5. Run the prompt again on the new diff until it would approve.

Prompt

Do a strict maintainability review of this branch's changes (git diff main...HEAD). Behaviour must stay the same. Judge structure only; I have other reviews for bugs and style.

Be ambitious. Before listing small fixes, ask: is there a restructuring ("code judo") that would make whole branches, flags, helpers or layers disappear? If so, lead with it.

Treat these as blockers unless there is a clear reason:
1. A file grows from under 1,000 lines to over 1,000 lines. Ask whether to split it first.
2. New if-statements, flags or special cases bolted into an unrelated or shared flow.
3. Feature logic leaking into a shared module, or logic placed in the wrong layer or package.
4. A new helper that duplicates one the codebase already has.
5. Thin wrappers or pass-through layers that add indirection and no clarity.
6. Casts, any, unknown or optional params that hide the real contract, or silent fallbacks covering an unclear rule.
7. Copy-pasted logic, or independent steps run one after another when that makes the flow more brittle.
8. Updates that can leave state half-applied when an atomic version is simple.

Output, in this order: structural regressions; missed big simplifications; branching and spaghetti growth; type and boundary problems; file size; smaller legibility points.
For each: file and line, the problem in one sentence, and the concrete restructure you propose.
Prefer a few high-conviction findings over many nits. Be direct but not rude.
End with APPROVE or CHANGES NEEDED and the blockers that decide it. Do not edit any code yet.

How to adapt

• Small scripts or a prototype? Raise or drop the 1,000-line rule.
• Not on git? Name the files or folders to review instead of the diff.
• Team has its own layering rules? Paste them under rule 3 so "wrong layer" is concrete.
• Want it to fix things too? After the review, reply "apply finding 1 only, then run the tests".

Gotchas

• Big restructures carry risk. Make sure tests cover the behaviour before you accept one.
• A strict reviewer can push abstraction for its own sake. Reject any suggestion that adds more concepts than it removes.
• Do not paste secrets, tokens or private customer data from config or logs into the chat with the diff.

Portability

Works with any language and any AI coding tool or chat AI. In a chat without repo access, paste the diff and the full files it touches instead of the git command.

Done when

• The review ends in APPROVE, or each remaining blocker has a written reason.
• No file crossed the size limit without a split or a reason.
• Tests still pass after every restructure you accepted.

Example prompt

"Run a strict maintainability review of this branch. Lead with any restructuring that deletes code, flag new special cases in shared code and files over 1,000 lines, and finish with approve or changes needed."

---

Adapted from "thermo-nuclear-code-quality-review" in cursor/plugins (cursor-team-kit, MIT, Copyright (c) 2026 Cursor). Original: https://github.com/cursor/plugins/blob/ae5650febf4320793844c12c4adaa937f417f8da/cursor-team-kit/skills/thermo-nuclear-code-quality-review/SKILL.md. See LICENSE.txt.

Bundled files

• LICENSE.txt

Licence

Adapted from Eric Zakariasson (Cursor, cursor/plugins cursor-team-kit, thermo-nuclear-code-quality-review) (Copyright (c) 2026 Cursor), MIT License. Adapted by TheSkillVault on 2026-11-03: the long rule list was condensed into one copy-paste prompt with a ranked output format and an approval bar; the Cursor-only frontmatter (disable-model-invocation) and the sample review phrases were dropped; the 1,000-line rule became adjustable; added steps, adapt notes, gotchas and done-when. Not the original file.

Licence: MIT License. Attribution when you reuse it: Based on work by Eric Zakariasson (Cursor, cursor/plugins cursor-team-kit, thermo-nuclear-code-quality-review) (Copyright (c) 2026 Cursor, MIT License), https://github.com/cursor/plugins/blob/ae5650febf4320793844c12c4adaa937f417f8da/cursor-team-kit/skills/thermo-nuclear-code-quality-review/SKILL.md; adapted by TheSkillVault, https://theskillvault.ai