Guide · Agent skill
How to use the Strix penetration testing agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Strix penetration testing skill teaches an AI agent how to run Strix, an open-source autonomous pentesting tool. Strix tries to exploit a target such as a web app, API, repository, URL, domain or IP, and it reports only findings it can back up with a working proof-of-concept. This guide explains what an agent skill is, how to install this one, how to ask for a scan in plain language, and what to check before trusting a result.
Strix Penetration Testingusestrix/strix · Apache-2.0 Download the skill Quick start
- 1Download and unzip the skill. Download the penetration-testing-with-strix ZIP from this page and unzip it so that SKILL.md sits in a penetration-testing-with-strix folder.
- 2Place it in your skills folder. Move the folder into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder, or upload the ZIP under Settings, Capabilities, Skills in the Claude apps.
- 3Install Docker and Strix. Confirm Docker is running with docker info, then install Strix with the curl script from the skill or with pipx install strix-agent, and check it with strix --version.
- 4Set your LLM environment variables. Export STRIX_LLM with a LiteLLM model id and LLM_API_KEY with your provider key. Do not commit these values to your repository.
- 5Ask for a scan in plain language. Request a pentest of a target you own, such as a staging URL or a local repo, and state any credentials or focus areas. The agent will choose the CLI or cloud and run the scan.
What an agent skill is and what this one makes the agent do
An agent skill is a folder with a SKILL.md file that tells an AI agent how to handle one kind of task. When a request matches the skill's description, the agent loads the instructions and follows them.
This skill covers security testing. It tells the agent how to choose between the self-hosted Strix CLI and the managed app.strix.ai cloud, how to start a scan with the right flags, and how to read the results. Strix looks for issues such as injection, XSS, SSRF, authentication and access-control flaws, IDOR and business logic bugs, and it aims to prove them rather than only flag them.
- Pentests a web app, API, website, repository, URL, domain or IP
- Runs headless scans with the open-source CLI or the managed cloud
- Returns validated findings as Markdown, JSON, CSV or SARIF
How to install the skill from the ZIP
Download the ZIP for penetration-testing-with-strix from this page and unzip it. The folder should contain SKILL.md.
For a personal install, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder so it applies only to that repository.
In the Claude apps, you can instead upload the ZIP under Settings, then Capabilities, then Skills.
- Personal: ~/.claude/skills/penetration-testing-with-strix/
- Project: .claude/skills/penetration-testing-with-strix/ inside your repo
- Claude apps: Settings, Capabilities, Skills, then upload the ZIP
Set up the Strix CLI, Docker and your LLM key
The self-hosted option needs three things. Docker must be running, because the CLI runs the agents in a Docker sandbox and the first scan pulls the sandbox image automatically. Strix must be installed, either with the install script from the source or with pipx. You also need an LLM provider API key and a model id, which the CLI reads from environment variables.
Never hardcode or commit your keys. Ask for the key only when it is not already set.
- Check Docker with docker info
- Check Strix with strix --version
- STRIX_LLM takes any LiteLLM model id, such as openai/... or anthropic/... or openrouter/...
- LLM_API_KEY holds your provider key
docker info
strix --version
export STRIX_LLM="<your LiteLLM model id>"
export LLM_API_KEY="<your provider api key>"Choose between the open-source CLI and the managed cloud
Neither option is better in general. The skill asks the agent to pick based on your situation. Use the open-source CLI when the source code must stay on your infrastructure, when you want a free local scan with Docker already present, when you use your own or a self-hosted model, or when your CI runner already has Docker and you want a self-contained gate.
Use the managed cloud when Docker is not available, when you do not want to manage an LLM key, when you want a team dashboard, scheduled scans, PR reviews or downloadable PDF/DOCX reports, or when the target is internal infrastructure your machine cannot reach. The managed cloud needs no Docker, no LLM key and no local compute. Its features are described in the managed-pentesting-with-strix skill.
- Source must stay local or the network is offline: CLI
- No Docker or no LLM key, or you want shared dashboards: cloud
- Both produce the same SARIF 2.1.0, so findings can be compared
Run a scan and read the results
Ask the agent for a scan in plain words, and it should start Strix in non-interactive mode. Headless mode is required for agents because the default interface blocks them. A scan can take minutes in quick mode or hours in deep mode, so the agent runs it in the background and checks back for completion.
Pass the target with -t. A local path is mounted into the sandbox writable, so point it at a clean checkout rather than uncommitted work. Always set a spend cap with --max-budget unless you have said otherwise.
Read results in strix_runs/<run-name>/. Start with penetration_test_report.md. Each validated finding has its own file in vulnerabilities/ with a proof-of-concept and remediation notes. The same findings are also available as vulnerabilities.json, vulnerabilities.csv and findings.sarif, and run.json records status, targets and cost.
- quick takes minutes, standard about 30 minutes, deep takes hours (the default)
- --fail-on critical makes the command exit 2 only for critical findings or worse
- --resume RUN_NAME continues a previous run from strix_runs/
strix -n -t https://staging.example.com --max-budget 20
strix -n -t ./ --scan-mode standard --max-budget 10What people use it for
- Scan a staging web app for IDOR and auth bypass before a release, using credentials you provide in the instruction.
- Run a headless SARIF scan in CI with --fail-on high so the pipeline fails only on serious findings.
- Test an OpenAPI spec together with its live API endpoint to cover both the description and the running service.
- Use the managed cloud for scheduled scans and PR reviews across a team, when the CLI on one laptop is not enough.
Frequently asked questions
Is the Strix penetration testing skill free to use?
The Strix open-source CLI is described as free and self-hosted, and the project is licensed under Apache-2.0. You still pay your LLM provider for the model calls, so set --max-budget to cap spend. The managed cloud is a separate paid service, and PDF/DOCX reports are listed as an Enterprise plan feature.
What do I need to install before using this skill?
For the self-hosted option you need Docker running, the Strix CLI installed via the curl script or pipx install strix-agent, and an LLM provider API key set as LLM_API_KEY with a model id in STRIX_LLM. The managed cloud needs none of these, only an app.strix.ai account.
Where do I put the skill folder for Claude?
Put the penetration-testing-with-strix folder in ~/.claude/skills/ for all your projects, or in .claude/skills/ inside one project. In the Claude apps you can upload the ZIP under Settings, Capabilities, Skills.
Can I use this skill without Docker?
Not for the open-source CLI, which runs scans in a Docker sandbox. The managed cloud runs on Strix infrastructure and needs no Docker, no LLM key and no local compute, so it is the option to pick in that case.
Why did my scan exit with code 0 but still miss things?
Exit code 0 means no validated vulnerabilities were found in what was analyzed, and it is not proof of full coverage. If the budget or turn limit is reached first, the scan wraps up early and still exits 0. Check strix_runs/<run>/run.json and the report's stated coverage, and rerun with a larger budget if needed.
Can I use Strix to test a site I do not own?
You should not. Strix actively exploits targets, which can cause damage or break services, so only test systems you own or have written permission to test.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits developers and security-minded teams who want an exploit-driven scan of an app, API or repository they own, either locally with Docker or through the managed cloud. It is a poor fit if you need a guaranteed complete audit, if you cannot run Docker or supply an LLM key and do not want cloud use, or if the target is not yours to test.
Download the skill