deepsafe-scan - Preflight security scanner for AI coding agents
Scans AI agent configs, skills, memory, and model behavior for security risks, secret leaks, PII, prompt injections, and dangerous patterns.
Tags
Updated: 2026-09-16Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Scan deployment configuration
- Scan skills and MCP servers
- Scan memory for secrets and PII
- Check hooks injection risks
- Probe model behavior safety
- Run LLM semantic analysis
- Generate security scan reports
Inputs
- Target scan directory
- Deployment configuration files
- Skill and MCP server files
- Session memory files
- AI agent configuration files
- Gateway URL and access token
Outputs
- JSON security scan report
- Markdown security scan summary
- HTML visual security report file
Requirements
- Python 3 runtime
- Read permission for target directory
- Bash execution permission
- Gateway URL and token for probes