From great-engineers
Dispatch one or more engineering personas to review existing code, architecture, or a draft spec. Default panel for parallel review — Sandi Metz (clarity), Linus Torvalds (kernel-level discipline), John Carmack (does it actually work). Override with --personas. Reads the project specification + the target file/directory + relevant adjacent code, produces a consolidated review with per-persona verdicts and a single highest-leverage recommendation. Output saves to engineering/reviews/<slug>.md.
How this skill is triggered — by the user, by Claude, or both
Slash command
/great-engineers:engineers-design-reviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Review existing code, architecture, or a draft spec.
Review existing code, architecture, or a draft spec.
Dispatches engineering personas to review the artifact at <path>. Default panel runs three personas in parallel (Sandi Metz, Torvalds, Carmack), each with a different lens. Output is a consolidated review — per-persona verdicts marked up, then a single recommended highest-leverage change.
Not for: line-level style nits (most projects have linters); auto-generated documentation; QA / test design (cross-dispatch great-minds:margaret-hamilton-qa instead).
When this skill is invoked with a <path> argument and optional --personas:
Resolve the project root the same way /engineers-project-init does. Verify CLAUDE.md exists or warn that the engineering bible is missing.
Read the project specification:
CLAUDE.md, README.md, the manifest, ARCHITECTURE.md, ADR/.great-authors/project.md if cross-craft projectRead the target. <path> may be:
src/auth/*.ts)Read the target file(s) in full. Read adjacent context — sibling files, the imports, the tests if present, any inline TODOs or FIXMEs. The reviewer can't review what they can't see.
Resolve the panel. If --personas is given (comma-separated), use those. Otherwise default panel:
sandi-metz-engineer — design clarity, refactor leverage, OO design qualitylinus-torvalds-engineer — kernel-level discipline, performance cost, what does this actually dojohn-carmack-engineer — minimum working solution, where does the time go, ship-readinessCommon alternative panels:
don-knuth-engineer + edsger-dijkstra-engineer — correctness and complexity proofanders-hejlsberg-engineer + sandi-metz-engineer — composability and claritybrendan-eich-engineer + linus-torvalds-engineer — backwards compat and performancesandi-metz-engineer + dhh-engineer — clarity vs pragmatism debategrace-hopper-engineer + sandi-metz-engineer — who can read this in five yearsDispatch the panel in parallel via the Agent tool. Each persona gets:
[→ replacement] for substitutions), Hand-off (if a different persona would serve better).Consolidate the parallel returns. The output is a single review file:
---
title: Design review of <path>
slug: <slug>
panel: [<persona-1>, <persona-2>, <persona-3>]
created: YYYY-MM-DD
target: <path>
target_lines: <N>
---
# Design review: <path>
## Per-persona verdicts
### <persona-1>
**Verdict:** <one sentence top-line reaction>
**Marked passages:**
- <quoted excerpt 1>
- <quoted excerpt 2>
- ...
**Hand-off:** <if a different persona would serve better; or omit>
### <persona-2>
[same structure]
### <persona-3>
[same structure]
## Where they agree
<1-3 points where the panel converges. The strongest signal — when independent reviewers with different lenses flag the same thing, it's almost certainly real.>
## Where they disagree
<1-2 points where the panel diverges. Often the most useful section — disagreement reveals the genuine trade-off the implementer or author is making.>
## Highest-leverage change
<ONE recommendation. The single change that, if made, would make the work meaningfully better. Not a list. The orchestrator's job is to pick the highest-leverage move; this skill picks it for them.>
## Suggested next step
<One of: implement the change, escalate to author, request a v2 from the spec author, run /engineers-debate <topic> <persona-A> <persona-B> if the disagreement is structural.>
Save the review to engineering/reviews/<slug>.md. Slug derived from the target path or from CLAUDE.md's Current spec: field.
Report:
📝 Saved to engineering/reviews/<slug>.md (review of <path>, <word-count> words).
Panel: <persona-1>, <persona-2>, <persona-3>
Convergence: <one-line summary of the agreement>
Highest-leverage change: <one-line summary>
Next:
- Address the highest-leverage change
- Or run /engineers-debate <topic> <a> <b> if the disagreement is structural
great-minds:margaret-hamilton-qa.)/engineers-design-review on every PR; use it for non-trivial changes where the consolidated lens is worth the dispatch cost.npx claudepluginhub sethshoultes/great-minds-constellation --plugin great-engineersProvides a checklist for code reviews covering functionality, security, performance, maintainability, tests, and quality. Use for pull requests, audits, team standards, and developer training.