==================================
Thermo-Nuclear Code Quality Review
==================================

Run an extremely strict maintainability review focused on abstraction
quality, giant files, and spaghetti-condition growth.

Description
-----------

The thermo-nuclear skill performs a deep code-quality audit of a branch
diff or user-specified paths. It prioritises "code judo": restructurings
that preserve behavior while making the implementation dramatically
simpler.

It is a **single-agent** review. It writes a report under
``resources/agent-docs/reviews/thermo-nuclear/`` and does not modify production source unless
the user later asks to implement remedies.

Triggers
--------

- ``thermo-nuclear``
- ``thermonuclear``
- ``thermo-nuclear code quality``
- ``deep code quality audit``
- ``harsh maintainability review``
- ``code judo review``

Invocation examples:

.. code-block:: text

   /wf:thermo-nuclear
   $thermo-nuclear
   /thermo-nuclear
   thermo-nuclear review of the current branch

Workflow
--------

1. **Resolve scope** — paths/PR if given; else current branch vs merge base.
2. **Gather evidence** — diff, surrounding code, file sizes near 1000 lines.
3. **Apply the rubric** — structural simplification, spaghetti growth,
   abstraction value, type/boundary cleanliness, canonical helpers,
   orchestration smells.
4. **Prioritise findings** — structural regressions and missed judo moves
   first; avoid nit floods.
5. **Apply the approval bar** — behavior-correct is not enough.
6. **Write report** — ``resources/agent-docs/reviews/thermo-nuclear/YYYY-MM-DD-HHMMSS.md``.
7. **Summarise** — verdict, blockers, top remedies.

Approval Bar (summary)
----------------------

Treat as presumptive blockers unless clearly justified:

- Missed dramatic simplification when a path is visible
- File pushed from under 1000 lines to over 1000 lines
- Ad-hoc branching that tangles an existing flow
- Feature checks scattered through shared code
- Unnecessary wrappers / cast-heavy contracts
- Logic in the wrong layer or duplicated canonical helpers

Output
------

- Session summary with verdict (PASS / WARN / FAIL)
- Markdown report under ``resources/agent-docs/reviews/thermo-nuclear/``

Relationship to Other Skills
----------------------------

- **team-review-impl** — multi-persona review against a written spec
- **complexity** — tool-based cyclomatic metrics; thermo-nuclear is
  qualitative structure review
- **review-tests** — test-only quality review

Attribution
-----------

Vendored and adapted from Cursor's ``thermo-nuclear-code-quality-review``
skill in `cursor/plugins <https://github.com/cursor/plugins>`_
(``cursor-team-kit``, MIT License).