security-scan

name	security-scan
description	One-shot security scan for the aiagent project. Checks for command injection in tool execution, path traversal in file tools, API key leakage, and MongoDB injection patterns. No iteration loop — immediate findings report.
user-invocable	true
argument-hint	scan internal/impl/tools/ for security issues
context	fork

Security Scan

Security scan for:

Instructions

Use the Agent tool with subagent_type: "security" to perform a one-shot security scan.

Pass it:

• The files or directory to scan, or a description of the feature to review
• Any specific concerns to focus on

The security agent will:

1. Run automated grep-based scans:
   • API key/secret patterns in code
   • Command injection risks (exec.Command usage)
   • Path traversal risks (file operations)
   • MongoDB filter injection patterns
   • Sensitive data in log statements
2. Review each flagged location in context
3. Propose specific remediations for any issues found

Output

The security agent produces:

• Automated scan results (what each grep found)
• Analysis of each finding (real vulnerability vs. false positive)
• Specific remediation for each real issue
• Summary: critical / warning / clean

Important

• This is a one-shot scan with no review loop — use /security-review for the full cycle
• Critical findings (injection vulnerabilities, key leakage) should be fixed before merging
• The scan covers the aiagent threat model: BashTool injection, FileReadTool/FileWriteTool traversal, API key logging, MongoDB NoSQL injection, WebSocket validation