pi-audit-master
Professional multi-agent codebase auditing and automated repair engine
Package details
Install pi-audit-master from npm and Pi will load the resources declared by the package manifest.
$ pi install npm:pi-audit-master- Package
pi-audit-master- Version
0.5.1- Published
- Jun 17, 2026
- Downloads
- 1,477/mo · 1,477/wk
- Author
- realvendex
- License
- unknown
- Types
- extension
- Size
- 52.6 KB
- Dependencies
- 0 dependencies · 1 peer
Pi manifest JSON
{
"extensions": [
"./extensions"
]
}Security note
Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.
README
🛡️ pi-audit-master
Professional Multi-Agent Codebase Auditing & Automated Repair Engine
pi-audit-master is a high-precision quality assurance tool for Pi extensions. Instead of generic code reviews, it deploys a coordinated fleet of specialized AI agents to identify critical bugs, logical gaps, and performance bottlenecks using industry-standard analysis techniques.
🚀 Core Features
🧬 The "Specialized Five" Audit Personas
The engine dispatches five parallel agents, each with a unique mental model:
| Persona | Focus | Technique |
|---|---|---|
| Type Sentinel | Null/Undefined Safety | Taint Analysis: Tracks data from source → flow → sink. |
| Logic Architect | Algorithmic Correctness | State-Machine Analysis: Finds race conditions and flow gaps. |
| Performance Oracle | Efficiency & Scaling | Complexity Analysis: Identifies O(n²) loops and leaks. |
| Ecosystem Integrator | Pi API Compatibility | Contract Analysis: Verifies Event and Factory patterns. |
| Quality Guardian | Maintainability | Smell Detection: Finds technical debt and redundant logic. |
⚙️ Advanced Capabilities
- Intelligent Mapping: The
ProjectMapperidentifies "Core Logic" files to maximize token efficiency and avoid auditing boilerplate. - Hybrid Reporting: Generates a professional
audit-report.mdfor the repository and a concise summary for the chat. - Automated Fix-Fleet: Optionally deploys a second wave of "Fixer" agents to resolve identified issues and verifies them via the project's test suite.
- Passive Mode: Automatically audits files after modifications (v0.4.0+)
- AI-Powered Analysis: Uses Pi's
complete()function for intelligent code review - Static Analysis Fallback: Works offline with pattern-based detection
- Timeout & Progress: Configurable timeouts and progress callbacks for long audits (v0.5.0+)
- Path Traversal Protection: Input sanitization to prevent path traversal attacks (v0.5.0+)
- Conservative Magic Number Detection: Avoids false positives on legitimate constants (v0.5.0+)
🛠️ Usage
Installation
pi install npm:pi-audit-master
Running an Audit
Invoke the tool via natural language or the direct command:
Natural Language: "Audit this project and fix any critical bugs."
Direct Command:
/audit {
"path": ".",
"depth": "deep",
"format": "hybrid",
"fix": true
}
Configuration Options
| Option | Type | Default | Description |
|---|---|---|---|
depth |
"surface" | "deep" |
"deep" |
Audit depth. Surface: entry points only. Deep: full core logic scan. |
format |
"chat" | "file" | "hybrid" |
"hybrid" |
Report format. Chat: summary only. File: markdown report. Hybrid: both. |
fix |
boolean |
false |
Enable automated Fix-Fleet to resolve issues. |
timeoutMs |
number |
300000 |
Optional timeout in milliseconds for the entire audit operation. |
onProgress |
(stage: string, progress: number, total: number) => void |
undefined |
Optional progress callback for UI updates. |
📈 Pipeline Flow
Command → Config → Project Mapping → Parallel Audit → Chief Synthesis → (Optional) Fix-Fleet → Verification.
📊 Version 0.5.0 Highlights
- All console output unified through logger utility — consistent formatting, configurable log levels
- Timeout mechanism — prevents hanging audits on large codebases
- Progress reporting — callbacks for UI integration
- Input sanitization —
sanitizePath()prevents path traversal - Conservative magic number fixes — avoids false positives on years, ports, HTTP codes, etc.
- TypeScript
isolatedModules: true— eliminated TS151002 warnings
📄 License
MIT