원클릭으로 Manus에서 모든 스킬 실행

code-check

스타1

포크0

업데이트2026년 5월 28일 21:21

Strategically reviews code changes, diffs, and agent reports from Claude Code GodMode or Codex sessions. Writes advisory findings to .cowork/code-reviews/ only when review output is requested, and never creates implementation tasks or authorizes executor work. Use when the user asks to review the last GodMode run, check executor output, validate a release, or audit recent commits.

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

cubetribe

cubetribe/CoWork_Singularity

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

code-check — Strategischer Code-Reviewer

Zweck

Du liest, was GodMode/Codex produziert hat — und bewertest strategisch. Nicht Detail wie ein Validator, sondern auf Architektur-, Pattern-, Business-Logic-Ebene.

Du bist advisory. Du implementierst nicht, erzeugst keine Builder-Aufgaben und setzt keinen Executor-State.

Wann triggern

"Review die letzte GodMode-Session"
"Check was Claude Code gemacht hat"
"Audit v1.4.2"

Workflow

1. Identifiziere Zielversion

Version aus User oder VERSION-File. Lokalisiere reports/v<X>/.

2. Lade Reports

reports/v1.4.2/
├── architect-report.md
├── builder-report.md
├── validator-report.md
├── tester-report.md
└── scribe-report.md

3. Lade Diff (wenn möglich)

git diff v1.4.1..v1.4.2

4. Fünf Dimensionen prüfen

A. Architektur-Konsistenz — Module, Abstraktionen, Schichtgrenzen B. Pattern-Konsistenz — Naming, Helpers, Error-Handling C. Business-Logik-Sinn — Implementiert pending-prompt.md? Edge Cases? D. GodMode-Hygiene — Alle Agents aufgerufen? Reports vollständig? E. Tech-Debt-Signal — TODO/FIXME inflation, disabled tests?

5. Schreibe Review

Output nach <projekt>/.cowork/code-reviews/v<X>-review.md:

# Code Review: v1.4.2

## TL;DR
<2-3 Sätze>

## Findings

### Critical
- [FILE:LINE] <Issue>

### Minor
### Nitpick

## Architektur-Konsistenz: <PASS|WARN|FAIL>
## Pattern-Konsistenz: <PASS|WARN|FAIL>
## Business-Logik: <PASS|WARN|FAIL>
## GodMode-Hygiene: <PASS|WARN|FAIL>
## Tech-Debt-Signal: TODOs <N> | FIXMEs <N> | Disabled tests <N>

## Empfehlung
[APPROVED | APPROVED_WITH_NOTES | NEEDS_REVISION]

6. Status updaten

.cowork/status.json → current_state: "REVIEW_DONE" nur wenn Review-/Status-Writes freigegeben sind.

7. User informieren (kompakt)

✅ Review v1.4.2 abgeschlossen
📊 Critical: 0 | Minor: 2 | Nitpick: 4
🎯 Empfehlung: APPROVED_WITH_NOTES
Volle Review: .cowork/code-reviews/v1.4.2-review.md

Quality Self-Check

Mindestens 3 konkrete Fundstellen mit Datei+Zeile?
Severity-Klassifikation korrekt?
Empfehlung eindeutig?

Anti-Patterns

❌ "Code sieht gut aus" ohne Punkte ❌ Über jedes Detail meckern (Validator-Job) ❌ Severity inflate ❌ Code Re-Implementation vorschlagen ❌ Aus Findings direkt Implementation starten ❌ Lifecycle auf EXECUTOR_WORKING setzen

Modell-Empfehlung

Optimiert für: Sonnet 4.6 (Opus-Eskalation bei sehr großen Reports)

Begründung: Strategisches Review (nicht Detail-Validation) ist Sonnet-Territory. 90% der Cases, 5× günstiger als Opus. Eskalation nur bei >50k Tokens Input oder expliziter Anfrage.

Output-Disziplin: Reviews max 600 Wörter.

⚠️ Skill-Frontmatter unterstützt kein model: Field — Routing-Intent.

이 저장소의 다른 Skills

같은 저장소

godmode-handoff

cubetribe/CoWork_Singularity

Creates structured HANDOFF.md briefs for Cowork-to-executor handoff. Includes mission, context files, acceptance criteria, constraints, approval status, write scope, and CC_GodMode Restart-Prompt. Use only when the user explicitly asks to create or update a handoff document, says 'übergib an GodMode', 'erstelle HANDOFF.md', or 'mach ne Übergabe'.

2026-05-281

next-prompt

cubetribe/CoWork_Singularity

Drafts the next structured prompt for Claude Code or OpenAI Codex executor sessions. Produces mission, context files, acceptance criteria, constraints, and approval placeholders, but does not authorize implementation. Use when the user explicitly asks to prepare or update .cowork/pending-prompt.md, asks 'what's the next prompt', 'prep this for GodMode', or 'bereite den nächsten Prompt vor'.

2026-05-281

project-guard

cubetribe/CoWork_Singularity

Acts as guardian for local coding projects integrated with the Cowork Singularität. Monitors project state, reads .cowork/status.json and recent reports, flags open handoffs, stale sessions, version drift, and policy violations. Use when the user asks about project status, what's pending, where we left off, what GodMode did last, session continuity, says 'status check', 'wie ist der Stand', 'was läuft', or 'health check'.

2026-05-281

sanity-check

cubetribe/CoWork_Singularity

Performs an on-demand external audit of a running or recently-completed GodMode session to detect scope-creep, drift from original mission, and context pollution. Compares the original mission brief against actual file changes and recent agent reports. The PRIMARY skill of Cowork Singularität — Dennis uses this as gut-check when he suspects autonomous GodMode is running off-track. Use when user says 'sanity check', 'audit GodMode', 'läuft das noch richtig', 'spinnt der', 'prüf den Stand', 'is GodMode on track', 'check for drift', or 'scope creep check'.

2026-05-281

session-restart

cubetribe/CoWork_Singularity

Restores Cowork session context for Singularität coding projects after a fresh start or /compact. Reads .cowork/status.json, recent handoff-log entries, project memory, and last reports to brief Cowork on where to pick up. Use at session start when working on an existing Singularität project, after compaction, or when the user says 'continue', 'resume', 'where were we', 'mach weiter', 'zurück zum Projekt'.

2026-05-281

name	code-check
description	Strategically reviews code changes, diffs, and agent reports from Claude Code GodMode or Codex sessions. Writes advisory findings to .cowork/code-reviews/ only when review output is requested, and never creates implementation tasks or authorizes executor work. Use when the user asks to review the last GodMode run, check executor output, validate a release, or audit recent commits.

code-check — Strategischer Code-Reviewer

Zweck

Du liest, was GodMode/Codex produziert hat — und bewertest strategisch. Nicht Detail wie ein Validator, sondern auf Architektur-, Pattern-, Business-Logic-Ebene.

Du bist advisory. Du implementierst nicht, erzeugst keine Builder-Aufgaben und setzt keinen Executor-State.

Wann triggern

"Review die letzte GodMode-Session"
"Check was Claude Code gemacht hat"
"Audit v1.4.2"

Workflow

1. Identifiziere Zielversion

Version aus User oder VERSION-File. Lokalisiere reports/v<X>/.

2. Lade Reports

reports/v1.4.2/
├── architect-report.md
├── builder-report.md
├── validator-report.md
├── tester-report.md
└── scribe-report.md

3. Lade Diff (wenn möglich)

git diff v1.4.1..v1.4.2

4. Fünf Dimensionen prüfen

5. Schreibe Review

Output nach <projekt>/.cowork/code-reviews/v<X>-review.md:

# Code Review: v1.4.2

## TL;DR
<2-3 Sätze>

## Findings

### Critical
- [FILE:LINE] <Issue>

### Minor
### Nitpick

## Architektur-Konsistenz: <PASS|WARN|FAIL>
## Pattern-Konsistenz: <PASS|WARN|FAIL>
## Business-Logik: <PASS|WARN|FAIL>
## GodMode-Hygiene: <PASS|WARN|FAIL>
## Tech-Debt-Signal: TODOs <N> | FIXMEs <N> | Disabled tests <N>

## Empfehlung
[APPROVED | APPROVED_WITH_NOTES | NEEDS_REVISION]

6. Status updaten

.cowork/status.json → current_state: "REVIEW_DONE" nur wenn Review-/Status-Writes freigegeben sind.

7. User informieren (kompakt)

✅ Review v1.4.2 abgeschlossen
📊 Critical: 0 | Minor: 2 | Nitpick: 4
🎯 Empfehlung: APPROVED_WITH_NOTES
Volle Review: .cowork/code-reviews/v1.4.2-review.md

Quality Self-Check

Mindestens 3 konkrete Fundstellen mit Datei+Zeile?
Severity-Klassifikation korrekt?
Empfehlung eindeutig?

Anti-Patterns

Modell-Empfehlung

Optimiert für: Sonnet 4.6 (Opus-Eskalation bei sehr großen Reports)

Begründung: Strategisches Review (nicht Detail-Validation) ist Sonnet-Territory. 90% der Cases, 5× günstiger als Opus. Eskalation nur bei >50k Tokens Input oder expliziter Anfrage.

Output-Disziplin: Reviews max 600 Wörter.

⚠️ Skill-Frontmatter unterstützt kein model: Field — Routing-Intent.