prompt-architect
Analyzes and transforms prompts using 7 research-backed frameworks (CO-STAR, RISEN, RISE-IE, RISE-IX, TIDD-EC, RTF, Chain of Thought, Chain of Density). Provides framework recommendations, asks targeted questions, and structures prompts for maximum effectiveness. Use when users need expert prompt engineering guidance.
When & Why to Use This Skill
Prompt-Architect is a high-performance Claude skill designed to transform vague user requirements into structured, research-backed prompts based on 2025 best practices. By leveraging seven advanced frameworks—including CO-STAR, RISEN, and Anthropic’s 'Contract-Style' templates—it automates the creation of explicit, context-aware instructions. The skill focuses on maximizing model adherence and minimizing hallucinations through 'Silent Planning,' XML-tag structuring, and integrated self-verification checklists, making it an essential tool for professional prompt engineering and agent development.
Use Cases
- Refining Vague Requirements: Automatically converting simple instructions like 'fix this bug' into comprehensive, role-based prompts with clear success criteria and 'No-Touch' zone constraints.
- Standardizing AI Workflows: Generating consistent 'Contract-Style' prompt templates for development teams to ensure all AI-generated outputs follow specific architectural patterns and coding standards.
- Reducing Model Hallucinations: Implementing 'Uncertainty Permission' and 'Self-Check' blocks that force the AI to verify its own output against constraints before responding.
- Context Optimization: Efficiently packaging project-specific context (such as CLAUDE.md or specific codebase excerpts) into prompts to stay within token budgets while providing sufficient information for complex tasks.
- Framework-Based Optimization: Recommending and applying the most suitable prompting framework (e.g., Chain of Thought for logic, Chain of Density for summarization) based on the specific nature of the user's task.
| name | prompt-architect |
|---|---|
| description | | |
| - Nate B. Jones 4 Beginner Moves | Shape, Context, Silent Plan, Self-Check |
| - Anthropic Claude 4.x Best Practices | Explizitheit, Contract-Style, Examples beat Adjectives |
Prompt-Architect: Best-Practice Prompt Generator
Basiert auf Claude 4.x Best Practices (Dezember 2025)
Quellen:
- Nate B. Jones Prompting Playbook 2025
- Anthropic Official Claude 4.x Prompting Guide
- Claude Code Best Practices
AKTIVIERUNG
Automatisch nach clarify-spec
Wenn clarify-spec ein JSON Output generiert hat, kann prompt-architect dieses JSON in einen strukturierten Prompt transformieren.
Manuell mit Trigger
/prompt-architect [Aufgabenbeschreibung]
Die 4 Beginner Moves (Nate B. Jones)
Move 1: Define the Shape of Output
NICHT: Schreib guten Code SONDERN: Produziere eine TypeScript-Datei mit Interface X, exportiere Funktion Y
Konkrete Artefakte definieren:
- Welche Dateien werden erstellt/geaendert?
- Welches Format hat der Output?
- Wie sieht ein Beispiel-Output aus?
Move 2: Give Minimal but Sufficient Context
- Automatisch relevante Dateien einlesen
- CLAUDE.md/AGENTS.md Auszuege
- Nur das Noetige - kein Context-Overflow
- Token Budget beachten
Move 3: Suggest a Silent Plan
- Claude soll INTERN planen bevor es antwortet
- Bei Claude 4.5: consider your approach first
- Vermeidung von think bei deaktiviertem Extended Thinking
Move 4: Add a Quick Self-Check
- Evaluator-Block am Ende des Prompts
- Checkliste: Format eingehalten? Constraints erfuellt?
- Before responding, verify: [checklist]
Claude 4.x Spezifische Patterns
Explizitheit + Modifier
NICHT: Create dashboard SONDERN: Create dashboard. Include as many relevant features as possible. Go beyond the basics.
Context + Motivation (WARUM erklaert WAS)
NICHT: NEVER use ellipses SONDERN: Your response will be read aloud by TTS, so never use ellipses since TTS cannot pronounce them.
Examples beat Adjectives
NICHT: Schreib professionell SONDERN: Beispiel-Output zeigen, Format demonstrieren
XML-Tags fuer Struktur
, ,
Uncertainty Permission
If unsure, state explicitly and ask clarifying question. Reduziert Halluzinationen drastisch.
Das Contract-Style Template
Nach Analyse des Auftrags, generiere diesen strukturierten Prompt:
---BEGIN GENERATED PROMPT---
ROLE and GOAL
You are: [Rolle basierend auf Task-Typ, z.B. Senior TypeScript Developer] Goal: [Konkrete Erfolgsdefinition in 1 Satz aus clarified_task.goal]
CONTEXT
[Automatisch gesammelt:]
- Projekt: [Projektname aus CLAUDE.md]
- Relevante Dateien: [Liste mit kurzem Inhaltsuebersicht]
- Bestehende Patterns: [Aus Codebase extrahiert]
- No-Touch Zones: [Aus clarified_task.scope.no_touch]
CONSTRAINTS
- [Constraint 1 als Bullet aus clarified_task.constraints]
- [Constraint 2 als Bullet]
- [Max 5 Constraints]
- If unsure about any aspect: State explicitly and ask for clarification
TASK
[Klare Aufgabenbeschreibung aus clarified_task.goal + problem_statement]
Success Criteria
- [Kriterium 1 aus clarified_task.success_criteria]
- [Kriterium 2]
- All existing tests must pass
Examples (if examples_needed = true)
Input: [Beispiel-Input] Expected Output: [Erwarteter Output]
OUTPUT FORMAT
[Exaktes Format basierend auf Task-Typ:] - Fuer Code: Vollstaendige Dateien, keine Diffs, keine Platzhalter - Fuer Docs: Markdown mit Headings - Fuer Config: JSON/YAML mit KommentarenVERIFICATION (Self-Check)
Before responding, verify: [ ] Output format followed exactly? [ ] All constraints from CONSTRAINTS section satisfied? [ ] All success criteria from TASK section met? [ ] Uncertain claims marked with [UNCERTAIN]? [ ] No changes to No-Touch Zones?
---END GENERATED PROMPT---
Workflow
Schritt 1: Input analysieren
Akzeptiert:
- JSON von clarify-spec (bevorzugt)
- Freie Textbeschreibung
- Kombiniert mit aktuellem Projekt-Kontext
Schritt 2: Kontext automatisch sammeln
- Glob fuer relevante Dateien
- Read CLAUDE.md/AGENTS.md
- Identify No-Touch Zones
- Extract existing patterns from codebase
Schritt 3: Template ausfuellen
Fuelle das Contract-Style Template mit:
- Role basierend auf Task-Typ (Frontend, Backend, DevOps, etc.)
- Goal aus clarified_task.goal
- Context aus Projekt-Analyse
- Constraints aus clarified_task.constraints
- Task aus goal + problem_statement
- Output Format basierend auf erwarteten Artefakten
- Verification Checklist
Schritt 4: Prompt praesentieren
Zeige den generierten Prompt dem User:
Generierter Best-Practice Prompt
[Vollstaendiger Prompt hier]
Optionen:
- ausfuehren: Diesen Prompt jetzt ausfuehren
- anpassen: Prompt modifizieren
- kopieren: Prompt in Zwischenablage
- speichern: Als Template speichern
Schritt 5: Ausfuehrung
Bei ausfuehren:
- Prompt intern verwenden
- Mit Quality Gates (wenn /supervisor aktiv)
- Ergebnis praesentieren
Beispiel: Vom vagen Auftrag zum Best-Practice Prompt
Input (von clarify-spec):
{ clarified_task: { goal: PPTX-Export fixen, damit AI-Bilder zuverlaessig erscheinen, problem_statement: Bilder fehlen manchmal im PPTX-Download, scope: { files: [presentationBuilder.ts, chartVisualGenerator.ts], no_touch: [api/analyze.ts] }, success_criteria: [Alle Bilder erscheinen, Unter 500KB, Keine Errors], constraints: [Keine Breaking Changes] } }
Output (generierter Prompt):
ROLE and GOAL
You are: Senior TypeScript Developer specializing in browser-based file generation Goal: Fix PPTX export so AI-generated images reliably appear in downloaded files
CONTEXT
- Projekt: fabrikIQ Analyzer (Manufacturing AI Platform)
- Relevante Dateien:
- presentationBuilder.ts: PptxGenJS integration, slide generation
- chartVisualGenerator.ts: Gemini image generation, base64 handling
- Bestehende Patterns: Async/await, try-catch error handling
- No-Touch Zones: api/analyze.ts (core analysis)
CONSTRAINTS
- No breaking changes to existing PPTX export functionality
- Images must be compressed to under 500KB
- Maintain existing error handling patterns
- If unsure about any aspect: State explicitly and ask
TASK
Fix the PPTX export to ensure AI-generated images reliably appear in downloads. The current issue: Images sometimes fail to appear in the final PPTX file.
Success Criteria
- All AI-generated images appear in PPTX
- Image file size under 500KB each
- No console errors during download
- Existing tests still pass
OUTPUT FORMAT
Provide complete, modified TypeScript files. No diffs, no placeholders, no partial implementations. Include brief comments explaining key changes.VERIFICATION
Before responding, verify: [ ] All images will be embedded correctly? [ ] Compression logic handles edge cases? [ ] No changes to api/analyze.ts? [ ] Error handling covers network failures?
Token-Effizienz
Der generierte Prompt ist optimiert fuer:
- Minimaler Context (nur relevante Infos)
- Maximale Klarheit (keine Ambiguitaet)
- Self-Verification (reduziert Iterationen)
Integration
Mit clarify-spec
clarify-spec -> JSON -> prompt-architect -> Ausfuehrung
Mit /supervisor
prompt-architect -> Prompt -> /supervisor -> Quality Gates -> Ergebnis
Standalone
/prompt-architect [Aufgabe] -> Prompt -> Ausfuehrung