Who it is forA security-minded developer who wants a first pass over a codebase without letting the agent change files.
The problemSecurity checks are often skipped during fast development because they take manual effort to set up.
The outcomeA review skill is installed and the developer has the guide's hooks and subagents modules to build a read-only audit flow.
What you need first
- Claude Code installed on macOS, Linux, or Windows
- git available to clone the guide
- A project folder where .claude/commands and CLAUDE.md can be created
Steps
git clone https://github.com/luongnv89/claude-howto.gitcp -r 03-skills/code-review-specialist ~/.claude/skills/- Read 04-subagents and 06-hooks in the guide to define a read-only audit subagent.
- Start a Claude Code session in the target project and ask it to audit for security issues without editing files.
Prompt for your agent
Using the repo luongnv89/claude-howto, install the code-review-specialist skill and use a subagent to audit <project-path> for security issues without modifying any file. Report the findings as a list with file paths. Verify that no file in <project-path> changed after the audit by comparing git status before and after.
Caveat. The README mentions read-only mode only as a use case label, so the exact read-only setup must be built from the subagents and hooks modules and checked by the user.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for luongnv89/claude-howto and "Run a read-only security audit with Claude Code subagents"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer