Pattern categories include XSS, SQL injection, command injection, unsafe code execution (eval/exec), SSRF, weak cryptography, hardcoded secrets, insecure deserialization, and path traversal/LFI/RFI, plus GraphQL-specific patterns (introspection, depth/complexity limiting, batching, authorization) when a GraphQL endpoint is present.

After scanning, it delegates fixes to the sniper agent with file:line, vulnerability, and fix — it does not apply fixes itself.

Security Scan Skill

Overview

Orchestrates the full security scanning workflow across all supported languages.

Supported Languages

Workflow

1. Detect language from project markers
2. Load patterns from references/scan-patterns.md
3. Run bun ${CLAUDE_PLUGIN_ROOT}/../node_modules/@fusengine/harness/dist/cli/bin.mjs scan <dir> for automated scanning (OWASP patterns ported into the harness)
4. Map findings to OWASP categories via references/owasp-top10.md
5. Generate report using references/templates/scan-report.md

Pattern Categories

• XSS (Cross-Site Scripting)
• SQL Injection
• Command Injection
• Code Execution (eval, exec)
• SSRF (Server-Side Request Forgery)
• Weak Cryptography
• Hardcoded Secrets
• Insecure Deserialization
• Path Traversal / LFI / RFI

Integration

After scanning, delegate fixes to sniper:

Agent(subagent_type="fuse-ai-pilot:sniper", prompt="Security fixes: [FILE:LINE] [VULN] [FIX]")

References

• OWASP Top 10 Mapping
• Scan Patterns by Language
• Report Template
• GraphQL Security Patterns — Load when the target exposes a GraphQL endpoint (introspection, depth/complexity limiting, batching, authorization checks).
• Scan Patterns - Python, Swift/iOS, Go, Rust — Load when scanning Python, Swift/iOS, Go, or Rust source code (patterns not covered in scan-patterns.md).