How to use the File Upload Vulnerability Checklist agent skill

What is an agent skill, and what does this one do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a conversation matches the skill's topic. This skill's SKILL.md contains a file upload vulnerability methodology that the agent uses as an operational checklist.

Once the skill is active and a conversation touches on upload testing, the agent is instructed to load the full methodology, follow the steps in order unless you specify otherwise, consider which techniques fit the current target, track which checklist items are completed, and suggest next steps based on findings.

• Mechanisms: an overview of how file upload flaws lead to issues such as remote code execution, client-side attacks, and denial of service.
• Extension testing: alternate web shell extensions, double extensions, case tricks, and trailing characters.
• Content-Type and magic byte testing: header manipulation and forged file headers.
• Polyglot files: GIFAR, EXIF-embedded PHP, PDF with PHP, and SVG with JavaScript.
• Path and filename abuse: traversal, UNC paths, and injection payloads inside filenames.
• Archive testing: Zip Slip, symlink abuse, and tar permission tricks.

Install the offensive-file-upload skill

Download the skill ZIP from this page and unzip it. The skill folder is named offensive-file-upload and contains SKILL.md. For a personal setup, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

You can also upload the ZIP in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload the skill there.

• Personal install: ~/.claude/skills/offensive-file-upload/
• Project install: .claude/skills/offensive-file-upload/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Trigger the skill with natural requests

The skill activates when your conversation involves its trigger phrases. These include file upload, MIME bypass, extension bypass, magic byte, path traversal upload, SVG XSS, polyglot, upload bypass, malicious upload, and web shell upload. You do not need to type these exactly; asking in plain language that mentions one of these topics is enough.

Give the agent context about the target and your authorization. The methodology asks the agent to consider applicability to the current target, so describing the upload feature helps it choose relevant techniques.

• "I'm running an authorized pentest. The profile picture upload accepts only JPEG and PNG. Which checks should I test first?"
• "Walk me through SVG XSS testing on our document upload endpoint, which is in scope."
• "The avatar upload seems to check Content-Type only. What MIME bypass techniques fit this?"
• "We've tested extension bypasses so far. What should we do next based on the results?"

What the agent will do during a test session

The skill follows a fixed workflow: load the methodology, work through the checklist in order, judge each technique against your target, track progress, and recommend next steps. Because the agent tracks completed items, you can return to a session and ask where you left off.

The methodology also lists testing prerequisites such as a collection of test files, an intercepting proxy like Burp Suite or ZAP, web shells for testing execution, and sandboxed converters for validation. Have these ready so the agent's suggestions can be acted on.

Safety notes: scope, scripts and network

The skill is offensive and is written for authorized penetration testing and bug bounty work. Only test systems you have written permission to test, or that are explicitly in scope for a bug bounty program. The source states that testing unauthorized systems is illegal.

The listed requirements are nothing, and the skill runs no scripts and makes no network calls. The skill is a set of instructions; any actual testing you perform happens through your own tools and targets.

• Use only against authorized, in-scope targets.
• The skill runs no scripts and uses no network access, according to its listing.
• Its requirements are not listed, so check your own tooling needs such as a proxy.
• It does not help you build secure upload handling.

Tips for getting useful results

Describe the upload function concretely: what file types are accepted, whether validation appears client-side or server-side, and any error messages you see. The methodology explicitly uses these indicators to decide which checks apply.

Work through the checklist in order unless you have a reason to skip ahead, and report findings back to the agent so it can suggest next steps.

• Share observed error messages and accepted types.
• Note whether validation seems to run in JavaScript or on the server.
• Ask the agent to summarize completed and remaining checklist items.