Agent System Audit Lite

A free, read-only self-check for AI workflow and agent-system readiness.

This tool scans a local project folder for basic operating layers that make AI-assisted work easier to resume, verify, and keep safe. It is intentionally small, local-only, and dependency-free.

What It Checks

The lite version checks six visible foundation layers:

• Project foundation: README and discoverable project entrypoints
• Basic secret hygiene: .gitignore env guards, no root .env, and obvious token-shaped text scan
• Tooling inventory: tools, bin, scripts, ops, or integration notes
• Verification basics: tests, smoke checks, CI, Makefile, or documented validation commands
• Documentation presence: docs, setup notes, architecture notes, or usage guide
• Automation structure: repeatable scripts/workflows instead of only manual process

What It Does Not Check

The paid human-reviewed audit covers deeper operational layers that usually need interpretation:

• Context doctrine and agent instructions
• Durable memory architecture
• Task/state persistence
• Observability, evidence, and run traces
• Handoff and multi-agent coordination
• Client-facing report quality
• Before/after improvement planning
• Human-reviewed prioritization and false-positive cleanup

Quick Start

python3 audit_lite.py --target /path/to/project

Write reports to a folder:

python3 audit_lite.py --target /path/to/project --out reports

The command writes:

• report.json
• report.md

Try The Examples

python3 audit_lite.py --target examples/healthy-workspace --out .tmp/healthy
python3 audit_lite.py --target examples/messy-workspace --out .tmp/messy

Example Report

A workspace with missing operating layers produces a report like this:

# Agent System Audit Lite Report

- Score: 2/9 (0.222, high risk)

## Layer Scores

- `automation-structure`: 0/1 (0.0)
- `basic-secret-hygiene`: 2/3 (0.667)
- `documentation`: 0/1 (0.0)
- `project-foundation`: 0/2 (0.0)
- `tooling-inventory`: 0/1 (0.0)
- `verification-basics`: 0/1 (0.0)

## Findings

- **FAIL** `readme-present` [project-foundation] README or project overview exists: 0/3 candidates found
  - Recommendation: Add a short README that explains what this project is, how to run it, and where key files live.
- **FAIL** `verification-path-present` [verification-basics] Tests, smoke checks, or validation commands exist: exit 1
  - Recommendation: Document at least one smoke test, lint, typecheck, build, or inspection command agents can run before claiming work is done.

Human-Reviewed Audit

Want a deeper review? This free tool checks basic agent readiness. For a human-reviewed audit with prioritized findings, risk notes, and a practical improvement roadmap for your AI workflow, agent system, or automation setup, book the Agent System Audit on Fiverr:

Book the Agent System Audit on Fiverr

Safety

• This tool is read-only.
• It does not call an AI model.
• It does not send data anywhere.
• It does not edit, install, delete, or upload files.
• It reports paths for secret-shaped findings, not secret values.
• It is not a security audit, compliance audit, or guarantee of production readiness.

Requirements

• Python 3.10+
• No third-party packages

License

MIT