Skip to content

sametbrr/prompt-architect

Repository files navigation

GitHub release License: MIT Agent Skills

Prompt Architect

An Agent Skill that turns any rough idea into a domain-classified, pattern-aware, quality-reviewed expert prompt. Supports input in Turkish and English.

🇹🇷 Türkçe için README.tr.md


Quick Start

git clone https://github.com/sametbrr/prompt-architect.git ~/.claude/skills/prompt-architect

Restart your Claude Code session, then trigger naturally:

> "Turn this into an expert prompt: build an onboarding strategy for a B2B SaaS"

Features

Given a vague input like "build an onboarding strategy" or "design an API for X", the skill:

  1. Analyzes the input — objective, constraints, complexity score (simple / moderate / complex).
  2. Classifies the domain against a 25-domain taxonomy with TR + EN signal keywords.
  3. Selects 3–6 prompting patterns (out of 9 UI-compatible ones) appropriate for the task.
  4. Drafts a refined English prompt using either a compact bullet scaffold or a full XML scaffold.
  5. Self-reviews the prompt against 8 quality gates.
  6. Executes the prompt and writes structured output to a file — only if the user asked for it.

The refined prompt body is always written in English for cross-model portability. Section labels follow the user's input language.


Requirements

  • Python 3.10+ (for the optional validator script — no pip install needed)
  • Claude Code or any agentskills.io-compatible agent

Installation

git clone https://github.com/sametbrr/prompt-architect.git ~/.claude/skills/prompt-architect

Claude Code auto-discovers skills under ~/.claude/skills/. Restart your session after cloning.


Usage

Modes

You want… Say something like… Mode
Just the refined prompt just the prompt, prompt only, don't run it prompt_only (default)
Prompt + the actual deliverable run it, execute it, generate the output too prompt_and_execute

Example

Input:

build an onboarding strategy for a B2B SaaS, and run it

Output (abbreviated):

Detected Domain: Product Growth Strategy
Complexity: moderate
Selected Patterns: Role, XML Structuring, Positive Guidance, CoT, Output Framing

Refined English Prompt:
<role>You are a senior product growth strategist...</role>
<task>Design a B2B SaaS onboarding strategy...</task>
...

Self-Review: 8/8 gates passed
Final Output: Saved to ./onboarding-strategy-output.md

Repository layout

prompt-architect/
├── SKILL.md                          # Skill entrypoint + 6-stage workflow
├── references/
│   ├── claude-prompting-patterns.md  # 9 UI-compatible patterns
│   ├── quality-gates.md              # 8 self-review gates + case study
│   ├── domain-taxonomy.md            # 25 domains, TR + EN signals
│   ├── mode-inference.md             # prompt_only vs prompt_and_execute
│   └── claude-md-rules.md            # Authoring rules applied dual-layer
├── assets/templates/
│   ├── refined-prompt-xml.tmpl       # Full XML scaffold (default for complex tasks)
│   ├── refined-prompt-compact.tmpl   # Bullet scaffold (simple tasks)
│   └── domain-*.tmpl                 # 6 domain packs
└── scripts/
    └── validate_prompt.py            # Optional 8-gate validator (stdlib only)

Key design choices

  • Purpose-built for Claude. Designed specifically for Claude Code, claude.ai, and Agent Skills-compatible clients, following Anthropic's official prompting practices end-to-end. Refined prompt outputs remain portable to other LLMs.
  • UI-compatible only. Patterns requiring API-level access (assistant prefill, thinking, stop_sequences, tool_choice) are out of scope. Everything works in a regular chat window.
  • Bilingual input, English body. The skill accepts Turkish or English input; the refined prompt body is always English for cross-model portability.
  • Minimum-viable patterns. 3–6 patterns per task, not all 9. Simplicity wins.
  • Don't over-ask. A clarifying question is allowed only if the input is genuinely unusable. Otherwise, infer and state assumptions.
  • File output discipline. In execute mode, structured deliverables (>~50 lines) are saved to disk; only summaries print inline.

Validator

python3 scripts/validate_prompt.py --stdin < draft-prompt.txt
python3 scripts/validate_prompt.py --self-test

Pure stdlib, Python 3.10+.


Compatibility

Tool Skills path Notes
Claude Code ~/.claude/skills/ or .claude/skills/ Global or project-level
GitHub Copilot (VS Code) .vscode/skills/ Agent mode required
OpenAI Codex ~/.codex/skills/ Same SKILL.md format
Cursor .cursor/skills/ Project-level
Gemini CLI ~/.gemini/skills/

License

MIT — see LICENSE.

About

Skill that turns any rough idea (TR/EN) into a domain-classified, quality-reviewed expert prompt.

Topics

Resources

License

Stars

Watchers

Forks

Contributors