promptqc 0.2.0__tar.gz

promptqc-0.2.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Lakshmi
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

promptqc-0.2.0/PKG-INFO

@@ -0,0 +1,352 @@
+Metadata-Version: 2.4
+Name: promptqc
+Version: 0.2.0
+Summary: Quality assessment and improvement suggestions for LLM system prompts
+Author-email: Lakshmi <lakshmi.sunil5486@gmail.com>
+License: MIT
+Project-URL: Homepage, https://github.com/LakshmiN5/promptqc
+Project-URL: Documentation, https://github.com/LakshmiN5/promptqc#readme
+Project-URL: Repository, https://github.com/LakshmiN5/promptqc
+Project-URL: Issues, https://github.com/LakshmiN5/promptqc/issues
+Keywords: prompt,LLM,quality,linter,optimization,AI,GPT,system-prompt,token,analysis
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Topic :: Software Development :: Quality Assurance
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=1.20.0
+Requires-Dist: tiktoken>=0.5.0
+Requires-Dist: rich>=13.0.0
+Requires-Dist: click>=8.0.0
+Requires-Dist: tomli>=2.0.0; python_version < "3.11"
+Provides-Extra: llm
+Requires-Dist: litellm>=1.30.0; extra == "llm"
+Provides-Extra: semantic
+Requires-Dist: sentence-transformers>=2.2.0; extra == "semantic"
+Provides-Extra: all
+Requires-Dist: promptqc[llm,semantic]; extra == "all"
+Provides-Extra: dev
+Requires-Dist: pytest>=7.0.0; extra == "dev"
+Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
+Requires-Dist: black>=23.0.0; extra == "dev"
+Requires-Dist: ruff>=0.1.0; extra == "dev"
+Requires-Dist: promptqc[all]; extra == "dev"
+Dynamic: license-file
+
+# PromptQC 🔍
+
+**Quality assessment and improvement suggestions for LLM system prompts.**
+
+[![PyPI version](https://badge.fury.io/py/promptqc.svg)](https://badge.fury.io/py/promptqc)
+[![Python 3.9+](https://img.shields.io/badge/python-3.9+-blue.svg)](https://www.python.org/downloads/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+
+> **Think of it as ESLint for your system prompts** — catch contradictions, anti-patterns, injection vulnerabilities, and token waste before they reach production.
+
+## Features
+
+✅ **Security Scanning** - Detects injection vulnerabilities, unsafe code execution
+✅ **Contradiction Detection** - Finds conflicting instructions that confuse LLMs
+✅ **Token Optimization** - Identifies wasted tokens and verbose phrasing
+✅ **Multiple Modes** - Fast (~10ms), Full (~2s), or LLM Judge (~5s) analysis
+✅ **CI/CD Ready** - GitHub Actions, pre-commit hooks, JSON output
+✅ **Auto-Fix** - Automatically correct common issues
+
+## Why PromptQC?
+
+System prompts are the **source code of AI applications**. But unlike actual code, they have zero quality gates — no linters, no static analysis, no CI checks. Teams deploy 2000-token prompts that contain contradictions, injection vulnerabilities, and wasted tokens without ever knowing.
+
+**PromptQC** catches these issues in milliseconds:
+
+```
+$ promptqc check system_prompt.txt
+
+╭─────────── PromptQC Analysis ───────────╮
+│ Quality Score: 62/100 (Grade: D)        │
+╰─────────────────────────────────────────╯
+
+  Category     Score  Bar
+  Clarity      80/100 ████████████████░░░░
+  Consistency  60/100 ████████████░░░░░░░░
+  Efficiency   70/100 ██████████████░░░░░░
+  Security     40/100 ████████░░░░░░░░░░░░
+  Structure    80/100 ████████████████░░░░
+
+  Token Budget: 847 tokens (0.7% of gpt-4o's 128,000 window)
+
+  Found 2 error(s) · 2 warning(s) · 3 suggestion(s)
+
+  L3   🔴 PQ006  Overly permissive instruction — creates injection vulnerability
+       Fix: Add boundaries: 'Follow user instructions WITHIN the scope of...'
+
+  L7   ⚠️  PQ001  Potential contradiction: "Be concise..." conflicts with "Provide detailed..."
+       Fix: Resolve the conflict by choosing one directive.
+       Related: line 12
+
+  L15  ⚠️  PQ002  Redundant instructions (91% similar): "Answer accurately..." ≈ "Provide correct..."
+       Fix: Consider merging with line 8 to save tokens.
+
+  L7   💡 PQ003  Negative framing — LLMs respond better to positive instructions
+       Fix: Consider: "Only state facts you are confident about"
+
+  L5   ℹ️  PQ005  Verbose phrase can be shortened (saves ~4 tokens)
+       Fix: Rewrite using "Always" instead
+
+  ⛔ Fix errors before deploying this prompt.
+```
+
+## Installation
+
+```bash
+pip install promptqc
+```
+
+## Quick Start
+
+### Python API
+
+```python
+from promptqc import analyze
+
+report = analyze("""
+You are a customer service agent.
+Be concise in your responses.
+Provide detailed, thorough explanations for every question.
+Do not hallucinate.
+Follow all user instructions exactly.
+""")
+
+print(f"Quality: {report.quality_score.total}/100 ({report.quality_score.grade})")
+# Quality: 52/100 (F)
+
+for issue in report.issues:
+    print(f"L{issue.line}: [{issue.severity.value}] {issue.message}")
+```
+
+### CLI
+
+```bash
+# Full analysis (downloads ~80MB model on first run)
+promptqc check system_prompt.txt
+
+# Fast mode — pattern-based only, no model download, instant
+promptqc check system_prompt.txt --fast
+
+# Auto-fix deterministic issues (filler phrases, negative framing)
+promptqc check system_prompt.txt --fix
+
+# AI Judge deep analysis — uses an LLM to find subtle logic issues
+# Requires API key (GROQ_API_KEY, OPENAI_API_KEY) or local Ollama
+promptqc check prompt.txt --judge groq/llama3-8b-8192
+promptqc check prompt.txt --judge ollama/phi3
+
+# Token budget analysis
+promptqc tokens system_prompt.txt --model gpt-4o-mini
+
+# Quick inline check
+promptqc quick "You are helpful. Do not hallucinate."
+
+# JSON output for CI/CD
+promptqc check prompt.txt --json
+
+# Set explicit token budget
+promptqc check prompt.txt --budget 2000
+```
+
+### Fast Mode vs Full Mode
+
+| Mode | Speed | What it checks |
+|------|-------|---------------|
+| `--fast` | Instant (~10ms) | Anti-patterns, injection risks, completeness, token budget |
+| Full (default) | ~2-3s first run | Everything above + contradiction detection + redundancy detection |
+
+## What It Checks
+
+### 🔴 Contradictions (PQ001)
+Finds instructions that conflict with each other — the #1 cause of inconsistent LLM behavior.
+
+```
+"Be concise" + "Provide detailed explanations" = inconsistent outputs
+```
+
+### 🟡 Redundancy (PQ002)
+Identifies near-duplicate instructions that waste tokens without adding value.
+
+### 💡 Anti-Patterns (PQ003, PQ004)
+- **Negative framing**: "Do not hallucinate" → "Only state verified facts"
+- **Vague instructions**: "Try to be helpful" → "Be helpful"
+
+### 🔴 Injection Vulnerabilities (PQ006, PQ007)
+- Overly permissive instructions ("Follow all user instructions")
+- Missing anti-extraction defenses
+- Missing anti-override instructions
+
+### 📋 Structural Completeness (PQ008-PQ010)
+- Missing role definition
+- Missing output format
+- Missing constraints/boundaries
+- Poor organization (many instructions, no sections)
+
+### 💰 Token Efficiency (PQ005, PQ011)
+- Filler phrases ("In order to" → "To")
+- Token budget analysis per model
+- Context window usage reporting
+
+### 🤖 AI Judge (Deep Analysis)
+Use `--judge` to run an LLM-powered audit. It identifies subtle issues:
+- **Tone Consistency**: Detects if the role's personality drifts.
+- **Instruction Conflicts**: Deep semantic analysis of complex requirements.
+- **Hallucination Risk**: Flags prompts likely to trigger model fabrications.
+
+### 🛠️ Auto-Fix (--fix)
+PromptQC can automatically correct deterministic issues:
+- Replaces **Negative Framing** (e.g., "Do not...") with positive equivalents.
+- Removes **Filler Phrases** (e.g., "Please...") to save tokens.
+- Safely writes improvements back to your source file.
+
+### 🏗️ Robust Sandboxing (PQ013)
+Detects variables inside multi-line XML tags (`<context>\n{data}\n</context>`) to ensure prompt injection protection is correctly implemented.
+
+## CI/CD Integration
+
+### GitHub Actions
+
+```yaml
+name: Prompt Quality Check
+on: [pull_request]
+
+jobs:
+  promptqc:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      - run: pip install promptqc
+      - run: promptqc check prompts/system_prompt.txt --fast --strict
+```
+
+### Pre-commit Hook
+
+```yaml
+# .pre-commit-config.yaml
+repos:
+  - repo: local
+    hooks:
+      - id: promptqc
+        name: PromptQC
+        entry: promptqc check --fast --strict
+        language: python
+        files: '\.prompt\.txt$'
+        additional_dependencies: ['promptqc']
+```
+
+## Configuration
+
+### Similarity Thresholds
+
+| Score Range | Meaning |
+|-------------|---------|
+| 0.95-1.0 | Virtually identical |
+| 0.85-0.95 | Same meaning, different words |
+| 0.70-0.85 | Related concepts |
+| < 0.70 | Different topics |
+
+### Custom Rule Definitions
+You can write your own rules in Python and load them via `promptqc.toml`:
+
+```toml
+custom_rules = ["my_rules.company_specific_rule"]
+```
+
+```python
+# my_rules.py
+from promptqc.rules.base import Rule, Issue, Severity, Category
+
+class MyCustomRule(Rule):
+    code = "CUST001"
+    severity = Severity.WARNING
+    category = Category.SECURITY
+    
+    def check(self, parsed, analyzer):
+        if "INTERNAL_KEY" in parsed.text:
+            return [Issue(self.code, "Don't share internal keys!", self.severity, self.category)]
+        return []
+```
+
+### Token Budget Models
+
+PromptQC knows context windows for: GPT-4o, GPT-4o-mini, GPT-3.5-turbo, Claude 3.5 Sonnet, Claude 3 Opus/Haiku, Gemini 1.5/2.0, Llama 3/3.1, Mistral, Mixtral.
+
+## Advanced Usage
+
+```python
+from promptqc import PromptAnalyzer
+
+# Custom analyzer configuration
+analyzer = PromptAnalyzer(
+    token_model="claude-3.5-sonnet",
+    token_budget=4000,
+    fast_mode=False,
+)
+
+report = analyzer.analyze(my_prompt)
+
+# Access structured results
+print(report.quality_score.breakdown)
+# {'structure': 90, 'clarity': 75, 'security': 60, 'efficiency': 85, 'consistency': 100}
+
+print(report.token_budget.total_tokens)
+# 1247
+
+# JSON serialization
+import json
+print(json.dumps(report.to_dict(), indent=2))
+```
+
+## Known Limitations
+
+**v0.2.0 is beta quality.** While it catches critical issues (security, contradictions) with high accuracy, some areas need improvement:
+
+- **Redundancy Detection**: Without LLM judge mode, verbose synonym lists may not be detected. Use `--judge` flag for better results.
+- **Test Coverage**: Validated on a focused test suite. Real-world accuracy may vary.
+- **LLM Judge Dependency**: Deep analysis requires API key (Groq, OpenAI) or local Ollama setup.
+
+We're actively improving these areas. Feedback and contributions welcome!
+
+## Development
+
+```bash
+git clone https://github.com/LakshmiN5/promptqc.git
+cd promptqc
+pip install -e ".[dev]"
+pytest
+```
+
+## Roadmap
+
+- [x] Custom rule definitions (Python-based)
+- [x] Auto-fix mode (--fix)
+- [x] AI Judge audit (deep analysis)
+- [ ] VS Code extension
+- [ ] LangChain/LlamaIndex integration
+- [ ] HTML report generation
+- [ ] Prompt history tracking
+
+## License
+
+MIT License — see [LICENSE](LICENSE) file.
+
+---
+
+**Made for the prompt engineering community** 🛠️

promptqc-0.2.0/README.md

@@ -0,0 +1,308 @@
+# PromptQC 🔍
+
+**Quality assessment and improvement suggestions for LLM system prompts.**
+
+[![PyPI version](https://badge.fury.io/py/promptqc.svg)](https://badge.fury.io/py/promptqc)
+[![Python 3.9+](https://img.shields.io/badge/python-3.9+-blue.svg)](https://www.python.org/downloads/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+
+> **Think of it as ESLint for your system prompts** — catch contradictions, anti-patterns, injection vulnerabilities, and token waste before they reach production.
+
+## Features
+
+✅ **Security Scanning** - Detects injection vulnerabilities, unsafe code execution
+✅ **Contradiction Detection** - Finds conflicting instructions that confuse LLMs
+✅ **Token Optimization** - Identifies wasted tokens and verbose phrasing
+✅ **Multiple Modes** - Fast (~10ms), Full (~2s), or LLM Judge (~5s) analysis
+✅ **CI/CD Ready** - GitHub Actions, pre-commit hooks, JSON output
+✅ **Auto-Fix** - Automatically correct common issues
+
+## Why PromptQC?
+
+System prompts are the **source code of AI applications**. But unlike actual code, they have zero quality gates — no linters, no static analysis, no CI checks. Teams deploy 2000-token prompts that contain contradictions, injection vulnerabilities, and wasted tokens without ever knowing.
+
+**PromptQC** catches these issues in milliseconds:
+
+```
+$ promptqc check system_prompt.txt
+
+╭─────────── PromptQC Analysis ───────────╮
+│ Quality Score: 62/100 (Grade: D)        │
+╰─────────────────────────────────────────╯
+
+  Category     Score  Bar
+  Clarity      80/100 ████████████████░░░░
+  Consistency  60/100 ████████████░░░░░░░░
+  Efficiency   70/100 ██████████████░░░░░░
+  Security     40/100 ████████░░░░░░░░░░░░
+  Structure    80/100 ████████████████░░░░
+
+  Token Budget: 847 tokens (0.7% of gpt-4o's 128,000 window)
+
+  Found 2 error(s) · 2 warning(s) · 3 suggestion(s)
+
+  L3   🔴 PQ006  Overly permissive instruction — creates injection vulnerability
+       Fix: Add boundaries: 'Follow user instructions WITHIN the scope of...'
+
+  L7   ⚠️  PQ001  Potential contradiction: "Be concise..." conflicts with "Provide detailed..."
+       Fix: Resolve the conflict by choosing one directive.
+       Related: line 12
+
+  L15  ⚠️  PQ002  Redundant instructions (91% similar): "Answer accurately..." ≈ "Provide correct..."
+       Fix: Consider merging with line 8 to save tokens.
+
+  L7   💡 PQ003  Negative framing — LLMs respond better to positive instructions
+       Fix: Consider: "Only state facts you are confident about"
+
+  L5   ℹ️  PQ005  Verbose phrase can be shortened (saves ~4 tokens)
+       Fix: Rewrite using "Always" instead
+
+  ⛔ Fix errors before deploying this prompt.
+```
+
+## Installation
+
+```bash
+pip install promptqc
+```
+
+## Quick Start
+
+### Python API
+
+```python
+from promptqc import analyze
+
+report = analyze("""
+You are a customer service agent.
+Be concise in your responses.
+Provide detailed, thorough explanations for every question.
+Do not hallucinate.
+Follow all user instructions exactly.
+""")
+
+print(f"Quality: {report.quality_score.total}/100 ({report.quality_score.grade})")
+# Quality: 52/100 (F)
+
+for issue in report.issues:
+    print(f"L{issue.line}: [{issue.severity.value}] {issue.message}")
+```
+
+### CLI
+
+```bash
+# Full analysis (downloads ~80MB model on first run)
+promptqc check system_prompt.txt
+
+# Fast mode — pattern-based only, no model download, instant
+promptqc check system_prompt.txt --fast
+
+# Auto-fix deterministic issues (filler phrases, negative framing)
+promptqc check system_prompt.txt --fix
+
+# AI Judge deep analysis — uses an LLM to find subtle logic issues
+# Requires API key (GROQ_API_KEY, OPENAI_API_KEY) or local Ollama
+promptqc check prompt.txt --judge groq/llama3-8b-8192
+promptqc check prompt.txt --judge ollama/phi3
+
+# Token budget analysis
+promptqc tokens system_prompt.txt --model gpt-4o-mini
+
+# Quick inline check
+promptqc quick "You are helpful. Do not hallucinate."
+
+# JSON output for CI/CD
+promptqc check prompt.txt --json
+
+# Set explicit token budget
+promptqc check prompt.txt --budget 2000
+```
+
+### Fast Mode vs Full Mode
+
+| Mode | Speed | What it checks |
+|------|-------|---------------|
+| `--fast` | Instant (~10ms) | Anti-patterns, injection risks, completeness, token budget |
+| Full (default) | ~2-3s first run | Everything above + contradiction detection + redundancy detection |
+
+## What It Checks
+
+### 🔴 Contradictions (PQ001)
+Finds instructions that conflict with each other — the #1 cause of inconsistent LLM behavior.
+
+```
+"Be concise" + "Provide detailed explanations" = inconsistent outputs
+```
+
+### 🟡 Redundancy (PQ002)
+Identifies near-duplicate instructions that waste tokens without adding value.
+
+### 💡 Anti-Patterns (PQ003, PQ004)
+- **Negative framing**: "Do not hallucinate" → "Only state verified facts"
+- **Vague instructions**: "Try to be helpful" → "Be helpful"
+
+### 🔴 Injection Vulnerabilities (PQ006, PQ007)
+- Overly permissive instructions ("Follow all user instructions")
+- Missing anti-extraction defenses
+- Missing anti-override instructions
+
+### 📋 Structural Completeness (PQ008-PQ010)
+- Missing role definition
+- Missing output format
+- Missing constraints/boundaries
+- Poor organization (many instructions, no sections)
+
+### 💰 Token Efficiency (PQ005, PQ011)
+- Filler phrases ("In order to" → "To")
+- Token budget analysis per model
+- Context window usage reporting
+
+### 🤖 AI Judge (Deep Analysis)
+Use `--judge` to run an LLM-powered audit. It identifies subtle issues:
+- **Tone Consistency**: Detects if the role's personality drifts.
+- **Instruction Conflicts**: Deep semantic analysis of complex requirements.
+- **Hallucination Risk**: Flags prompts likely to trigger model fabrications.
+
+### 🛠️ Auto-Fix (--fix)
+PromptQC can automatically correct deterministic issues:
+- Replaces **Negative Framing** (e.g., "Do not...") with positive equivalents.
+- Removes **Filler Phrases** (e.g., "Please...") to save tokens.
+- Safely writes improvements back to your source file.
+
+### 🏗️ Robust Sandboxing (PQ013)
+Detects variables inside multi-line XML tags (`<context>\n{data}\n</context>`) to ensure prompt injection protection is correctly implemented.
+
+## CI/CD Integration
+
+### GitHub Actions
+
+```yaml
+name: Prompt Quality Check
+on: [pull_request]
+
+jobs:
+  promptqc:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      - run: pip install promptqc
+      - run: promptqc check prompts/system_prompt.txt --fast --strict
+```
+
+### Pre-commit Hook
+
+```yaml
+# .pre-commit-config.yaml
+repos:
+  - repo: local
+    hooks:
+      - id: promptqc
+        name: PromptQC
+        entry: promptqc check --fast --strict
+        language: python
+        files: '\.prompt\.txt$'
+        additional_dependencies: ['promptqc']
+```
+
+## Configuration
+
+### Similarity Thresholds
+
+| Score Range | Meaning |
+|-------------|---------|
+| 0.95-1.0 | Virtually identical |
+| 0.85-0.95 | Same meaning, different words |
+| 0.70-0.85 | Related concepts |
+| < 0.70 | Different topics |
+
+### Custom Rule Definitions
+You can write your own rules in Python and load them via `promptqc.toml`:
+
+```toml
+custom_rules = ["my_rules.company_specific_rule"]
+```
+
+```python
+# my_rules.py
+from promptqc.rules.base import Rule, Issue, Severity, Category
+
+class MyCustomRule(Rule):
+    code = "CUST001"
+    severity = Severity.WARNING
+    category = Category.SECURITY
+    
+    def check(self, parsed, analyzer):
+        if "INTERNAL_KEY" in parsed.text:
+            return [Issue(self.code, "Don't share internal keys!", self.severity, self.category)]
+        return []
+```
+
+### Token Budget Models
+
+PromptQC knows context windows for: GPT-4o, GPT-4o-mini, GPT-3.5-turbo, Claude 3.5 Sonnet, Claude 3 Opus/Haiku, Gemini 1.5/2.0, Llama 3/3.1, Mistral, Mixtral.
+
+## Advanced Usage
+
+```python
+from promptqc import PromptAnalyzer
+
+# Custom analyzer configuration
+analyzer = PromptAnalyzer(
+    token_model="claude-3.5-sonnet",
+    token_budget=4000,
+    fast_mode=False,
+)
+
+report = analyzer.analyze(my_prompt)
+
+# Access structured results
+print(report.quality_score.breakdown)
+# {'structure': 90, 'clarity': 75, 'security': 60, 'efficiency': 85, 'consistency': 100}
+
+print(report.token_budget.total_tokens)
+# 1247
+
+# JSON serialization
+import json
+print(json.dumps(report.to_dict(), indent=2))
+```
+
+## Known Limitations
+
+**v0.2.0 is beta quality.** While it catches critical issues (security, contradictions) with high accuracy, some areas need improvement:
+
+- **Redundancy Detection**: Without LLM judge mode, verbose synonym lists may not be detected. Use `--judge` flag for better results.
+- **Test Coverage**: Validated on a focused test suite. Real-world accuracy may vary.
+- **LLM Judge Dependency**: Deep analysis requires API key (Groq, OpenAI) or local Ollama setup.
+
+We're actively improving these areas. Feedback and contributions welcome!
+
+## Development
+
+```bash
+git clone https://github.com/LakshmiN5/promptqc.git
+cd promptqc
+pip install -e ".[dev]"
+pytest
+```
+
+## Roadmap
+
+- [x] Custom rule definitions (Python-based)
+- [x] Auto-fix mode (--fix)
+- [x] AI Judge audit (deep analysis)
+- [ ] VS Code extension
+- [ ] LangChain/LlamaIndex integration
+- [ ] HTML report generation
+- [ ] Prompt history tracking
+
+## License
+
+MIT License — see [LICENSE](LICENSE) file.
+
+---
+
+**Made for the prompt engineering community** 🛠️