Code Review Expert Suite
The Code Review Expert suite provides deep, rigorous code review tools. It includes two primary modes of operation: the autonomous parallel code-review-expert review, and the interactive mentor-style code-review-adversary review.
1. Code Review Expert Parallel Audit (Autonomous)
Launch the two thermo review passes as async background subagents in parallel, then synthesize results.
Workflow
- Determine review scope from user request, PR, current branch, or changed files.
- Gather diff + file context needed for reviewers.
- Launch both subagents in parallel:
thermo-nuclear-review-subagent — bugs, breakages, security, devex regressions, feature-flag leaksthermo-nuclear-code-quality-review-subagent — maintainability, structure, file-size, spaghetti, abstractions
- Pass each the same scoped diff/context.
- After both finish, synthesize: findings first, deduplicated, overlapping findings weighted more heavily.
2. Code Review Adversary (Interactive Mentoring)
Use code-review-adversary when the user wants a strict, expert-grade review framed as a Senior Engineer reviewing a Junior Developer. It offers interactive configuration (model, criticality, inline PR comments) and focuses on teaching the "why" behind every finding.
Skills Included
| Skill |
Description |
thermo-nuclear-review |
Deep branch audit (bugs, breakages, security, devex, feature-flag leaks) |
thermo-nuclear-code-quality-review |
Strict maintainability audit (code-judo, 1k-line rule, spaghetti, boundaries) |
code-review-expert |
Run both review subagents in parallel and synthesize |
code-review-adversary |
Adversarial code review where the reviewer plays a senior software engineer teaching a junior developer |
Usage
For an autonomous double-pass audit (code-review-expert):
Gather git diff main...HEAD and full contents of changed files. Invoke both thermo-nuclear-* subagents in one message with run_in_background: true. Synthesize prioritized, deduped findings.
For a mentoring review (code-review-adversary):
Gather the diff and prompt the user interactively (Scope, Model, Criticality, Inline comments) if not already specified, then generate a single comprehensive report or inline PR comments.
1---2name: code-review-expert3description: Run both thermo-nuclear review (security/correctness) and thermo-nuclear-code-quality-review (maintainability/structure) subagents in parallel, then synthesize findings. Use for full-spectrum branch audits combining bug/security and code-quality passes. The plugin also includes `code-review-adversary` for interactive, mentor-style senior-vs-junior PR reviews.4---56# Code Review Expert Suite78The Code Review Expert suite provides deep, rigorous code review tools. It includes two primary modes of operation: the autonomous parallel `code-review-expert` review, and the interactive mentor-style `code-review-adversary` review.910## 1. Code Review Expert Parallel Audit (Autonomous)1112Launch the two thermo review passes as async background subagents in parallel, then synthesize results.1314### Workflow151. Determine review scope from user request, PR, current branch, or changed files.162. Gather diff + file context needed for reviewers.173. Launch both subagents in parallel:18 - `thermo-nuclear-review-subagent` — bugs, breakages, security, devex regressions, feature-flag leaks19 - `thermo-nuclear-code-quality-review-subagent` — maintainability, structure, file-size, spaghetti, abstractions204. Pass each the same scoped diff/context.215. After both finish, synthesize: findings first, deduplicated, overlapping findings weighted more heavily.2223## 2. Code Review Adversary (Interactive Mentoring)2425Use `code-review-adversary` when the user wants a strict, expert-grade review framed as a Senior Engineer reviewing a Junior Developer. It offers interactive configuration (model, criticality, inline PR comments) and focuses on teaching the "why" behind every finding.2627## Skills Included2829| Skill | Description |30|-------|-------------|31| `thermo-nuclear-review` | Deep branch audit (bugs, breakages, security, devex, feature-flag leaks) |32| `thermo-nuclear-code-quality-review` | Strict maintainability audit (code-judo, 1k-line rule, spaghetti, boundaries) |33| `code-review-expert` | Run both review subagents in parallel and synthesize |34| `code-review-adversary` | Adversarial code review where the reviewer plays a senior software engineer teaching a junior developer |3536## Usage3738**For an autonomous double-pass audit (`code-review-expert`):**39Gather git diff main...HEAD and full contents of changed files. Invoke both `thermo-nuclear-*` subagents in one message with `run_in_background: true`. Synthesize prioritized, deduped findings.4041**For a mentoring review (`code-review-adversary`):**42Gather the diff and prompt the user interactively (Scope, Model, Criticality, Inline comments) if not already specified, then generate a single comprehensive report or inline PR comments.
Andersonlimahw