메뉴
Two-axis review of a diff — Standards (does it follow the repo's conventions and avoid code smells?) and Spec (does it implement what the ticket/spec asked?). Runs the axes as parallel sub-agents. Use to review a branch, PR, or work-in-progress.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Relentlessly interrogate a plan or design one question at a time until the goal, risks, and trade-offs are clear.
Implement one ready ticket or spec interactively in the foreground — the attended alternative to the AFK /nightmanager night shift. TDD where seams allow, then code review, then commit.
Investigate a question against high-trust primary sources and capture the findings as a Markdown note in the repo. Dispatches Nightmanager's librarian tool so you keep working while it reads.
Test-driven development reference — what a good test is, where tests go, the anti-patterns, and the rules of the red → green loop. Reference material for /implement and the night shift, not a workflow of its own.
Turn the current conversation into a draft local spec in specs/ using this repository's spec template and Nightmanager conventions. Synthesis, not an interview.
Break a spec, plan, or conversation into vertically-sliced Nightmanager tickets — local TODOs.md entries, each declaring what it delivers and what blocks it.
| name | code-review |
| description | Two-axis review of a diff — Standards (does it follow the repo's conventions and avoid code smells?) and Spec (does it implement what the ticket/spec asked?). Runs the axes as parallel sub-agents. Use to review a branch, PR, or work-in-progress. |
Review the diff between HEAD and a fixed point along two independent axes:
The two axes run as parallel sub-agents so they don't pollute each other's context; this skill then reports them side by side. A change can pass one and fail the other (right thing built wrong, or wrong thing built cleanly), so the axes are never merged or re-ranked.
The shared criteria live in Nightmanager's prompts/review-personas.md — the single source of truth for the review lenses and the code-smell baseline. This skill orchestrates; that prompt defines what "good" means.
Use whatever fixed point the user gives (a SHA, branch, tag, main, HEAD~5). If none, ask. Capture git diff <fixed-point>...HEAD (three-dot, against the merge-base) and git log <fixed-point>..HEAD --oneline. Confirm the ref resolves and the diff is non-empty before spawning sub-agents — a bad ref or empty diff should fail here.
Find the originating spec, in order: issue references in the commit messages; a path the user passed; a spec under specs/ matching the branch/feature; else ask. If there is genuinely no spec, the Spec sub-agent skips and reports "no spec available".
Send one message with two general-purpose (or oracle) sub-agent calls.
Standards sub-agent — give it the diff command, the commit list, the repo's documented standards (if any), and the shared review lenses + Fowler smell baseline from prompts/review-personas.md, pasted in full (the sub-agent has no other access to it). Brief: "Report per file/hunk — (a) each place the diff breaks a documented repo standard (cite it); (b) any smell from the baseline (name it, quote the hunk). Documented repo standards are hard calls and override the baseline; baseline smells are always judgement calls. Skip anything tooling already enforces. Under 400 words."
Spec sub-agent — give it the diff command, commit list, and the spec contents. Brief: "Report — (a) spec requirements missing or partial; (b) behaviour not asked for (scope creep); (c) requirements implemented but implemented wrong. Quote the spec line for each finding. Under 400 words."
Present the reports under ## Standards and ## Spec, verbatim or lightly cleaned. Do not merge or re-rank across axes. End with a one-line summary: findings per axis and the worst issue within each axis — no single cross-axis "winner".
/nightmanager applies the same review lenses inline before it commits (embedded in prompts/nightmanager.md, mirrored from prompts/review-personas.md). This skill is the on-demand, foreground equivalent used by /implement and by humans reviewing a branch or PR.
