claude-pentest plugin
A Claude Code plugin with 15 agents and 63 attack categories for authorized penetration testing with sign-off.
- 133 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Executor agents run active exploitation (injection, CSP bypass, CVE testing) against a real target; running this against a system without authorization has legal consequences.
- The optional Kali server gives access to Metasploit and other post-exploitation tooling.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/claude-pentestDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Executor agents run active exploitation (injection, CSP bypass, CVE testing) against a real target; running this against a system without authorization has legal consequences.
- The optional Kali server gives access to Metasploit and other post-exploitation tooling.
This is third-party code. Review the repository files before installing.
What it does
claude-pentest turns Claude Code into a pentest engagement coordinator: the `/pentest:pentest` command collects scope and an attack profile, runs reconnaissance, then asks a planner agent for a deployment plan and shows it for approval. Once approved, specialized executor agents run in parallel under a 4-phase protocol: injection, XSS, CSRF, CSP bypass, prototype pollution, CVE testing and more across 63 categories in 11 domains. Every finding ships with a working PoC, captured HTTP evidence and a Playwright screenshot rather than a theoretical description. For server-side tools (nmap, sqlmap, gobuster, Metasploit) a remote Kali server connects through `/pentest:pentest-kali`; without it, agents fall back to whatever is available locally via Bash.
Who it is for. Pentesters and security engineers running authorized tests of their own or client systems who want a coordinated agent set instead of running each tool by hand.
Good fit when
- A new pentest engagement is starting, and scope, attack profile and Kali server need collecting in one conversation, not by hand across three files.
- Recon already produced an endpoint list, and injection, XSS and auth checks need to run in parallel with evidence.
- A potential CVE has turned up, and an executor should take it to a working PoC with captured traffic.
- The engagement is already underway and only the attack profile needs updating, without re-collecting the scope.
Not a fit when
- There is no written permission to test the target system: the README explicitly asks for it upfront.
- A quick single scanner without agent coordination or a pre-exploitation approval step is what's needed.
- No Kali server is configured and local tools (nmap, sqlmap and similar) are not installed on the machine: several attack categories go uncovered.
Example request
Run /pentest:pentest against this staging target, web application profile, medium thoroughness, markdown report.Limitations
Server-side tools (nmap, sqlmap, gobuster, Metasploit) need a separate network-reachable Kali Linux box running MCP-Kali-Server; without it, agents use only what is available locally via Bash. The `/pentest:pentest` command itself dispatches parallel executors from the main session, because the Claude Code harness allows only one level of subagent nesting. Exploitation starts only after the operator approves the plan, but recon and scanning run automatically right after launch. Optional session isolation to the plugin is lifted with `/pentest:pentest-exit` or `/clear`.
How to disable. Remove the plugin with `/plugin uninstall pentest@claude-pentest`. Delete the saved scope (`.pentest-scope.json`), attack profile (`.pentest-attacks.json`) and Kali settings (`.pentest-mks.json`) files from the project separately.
FAQ
How do I install claude-pentest in Claude Code?
Add the marketplace first: `/plugin marketplace add Stickman230/claude-pentest`, then `/plugin install pentest@claude-pentest`. All agents and commands become available in any project session.
Is a Kali server required?
No. Without one, agents use whatever is available locally via Bash. With a server (via `/pentest:pentest-kali`, checked over `/health`) nmap, sqlmap, gobuster, dirb, nikto, hydra, john and Metasploit become available.
Does the plugin start attacking without approval?
Recon and plan building run automatically after launch, but active exploitation needs an explicit operator approval of the plan at the step before executor agents deploy.
How is claude-pentest different from a single scanner?
It is a coordination layer: a planner decides which of 63 attack categories across 11 domains to run and in what order, rather than one fixed set of checks.
What does an engagement produce?
A structured `pentest-report.json` and `report.md` under `outputs/{engagement}/`, with PoC files, captured HTTP traffic and screenshots for each finding.
Discussion
No questions or reviews yet.
Related
PluginsA pack of about fifty Claude Code subagents for authorized pentesting: planning, recon, exploit research and chaining, post-exploitation, detection and reportin
An authorized bug-bounty suite: 50 subagents, 26 commands and 2 bug-bounty-platform MCP servers for Claude Code, Codex and five more agents
Ten pentest skills for Claude Code and Gemini CLI: recon, SQL injection, XSS, LFI and a report, driven by plain-language requests
A CLI agent for authorized pentesting: from a natural-language brief it runs recon, finds and verifies vulnerabilities and prepares a report
hackingtool plugin
hackingtool - Claude Code plugin
A Claude Code plugin wrapping 183 pentest and OSINT tools from Z4nzu/hackingtool, running them locally via Bash, WSL or Docker
Self-hosted AI pentesting framework: attack surface graph, Kali sandbox, human approval gates and an MCP server you can drive from Claude Code
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| claude-pentest (this card) | no | 133 | ||
| Pentest Agent Suite | no | key is optional | 1,057 | a different agent set with no dedicated planner or 4-phase protocol per executor |
| MCP Kali Server | no | not required | 871 | an MCP server for Kali itself; claude-pentest connects to it as an optional backend rather than replacing it |
README in short
The README, in English, French and Simplified Chinese, walks through all five slash commands and their step flow, with a dedicated section on the MCP-Kali-Server integration.