Attune Lite

Six developer workflow commands for Claude Code: code review, security audit, test generation, test fixing, bug prediction, and documentation generation. No dependencies required.

Install

Step 1: Add the marketplace:

claude plugin marketplace add Smart-AI-Memory/attune-lite

Step 2: Install the plugin:

claude plugin install attune-lite

Verify it loaded:

/plugins

Commands

Command	What it does
/code-review <path>	Multi-pass code review (security, quality, performance, architecture)
/security-audit <path>	OWASP-aligned vulnerability scan with remediation plan
/smart-test <path>	Find test gaps and generate tests for uncovered code
/fix-test --lf	Auto-diagnose and fix failing tests with retry loop (up to 3 attempts)
/bug-predict <path>	Predict likely bug locations from patterns and complexity
/doc-gen <path>	Generate documentation from source code

All commands default to the current working directory if no path is provided.

How It Works

Each command triggers a structured, multi-pass analysis skill. For example, /code-review runs four passes (security, quality, performance, architecture) and synthesizes findings into a single report with scores, file references, and prioritized fixes.

Skills are prompt-based -- they guide Claude through a systematic analysis process. No external tools, APIs, or package installations are needed.

Example

/code-review src/

Produces a report like:

## Summary
Score: 82/100 | Files reviewed: 24 | Issues: 11

## Security
| Severity | File       | Line | Issue              | Fix           |
|----------|------------|------|--------------------|---------------|
| HIGH     | auth.py    | 45   | Missing input validation | Add schema check |
| MEDIUM   | config.py  | 12   | Hardcoded secret   | Use env var   |

## Top Priorities
1. Add input validation to auth.py:45
2. Move secrets to environment variables
3. Add error handling to db.py:88

When to Use

• Before opening a PR -- run /code-review to catch issues early
• During security reviews -- run /security-audit for a structured vulnerability assessment
• When adding features -- run /smart-test to generate tests for new code
• When tests fail -- run /fix-test --lf to auto-diagnose and fix failures instead of manually reading each traceback
• Before releases -- run /bug-predict to find high-risk areas
• When onboarding -- run /doc-gen to generate documentation for unfamiliar code

Troubleshooting

Command not recognized: Make sure the plugin is installed. Run /plugins to verify attune-lite appears in the list.

Review seems incomplete: For large codebases, specify a narrower path (e.g., /code-review src/auth/ instead of /code-review .).

About

Attune Lite surfaces the best workflow patterns from attune-ai as native Claude Code skills. For the full framework with 15 workflows, CLI integration, and multi-agent orchestration:

pip install 'attune-ai[developer]'

License

Apache-2.0