XiaoYiWeio

DeepSafe Scan — Claude Code Integration

Security community intermediate

SOURCE https://github.com/XiaoYiWeio/deepsafe-scan

Description

This directory contains **deepsafe-scan**, a preflight security scanner for AI agent environments.

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

Repository README

This is the README for XiaoYiWeio/deepsafe-scan, shared by 3 entries in this directory. It describes the repository, not this entry specifically.

DeepSafe Scan — Claude Code Integration

This directory contains **deepsafe-scan**, a preflight security scanner for AI agent environments.

What it does

Scans for security risks across 5 dimensions:

Module	What it checks	Needs API?
posture	openclaw.json / .env config security	No
skill	Installed skills & MCP servers — 15+ static analyzers	No
memory	Session data — 27+ secret patterns, 9 PII patterns	No
hooks	`.claude/settings.json`, `.cursorrules`, `.vscode/tasks.json` — 12 injection patterns	No
model	4 behavioral safety probes (persuasion, sandbagging, deception, hallucination)	Yes (ANTHROPIC_API_KEY)

When to use

Invoke this scanner when the user asks to:

"Check if this project's AI config is safe"
"Scan my setup for security issues"
"Audit my skills/MCP servers"
"Check for prompt injection in agent configs"
"Run a security scan before I start using this project"

How to run

# Quick static scan (no API key needed) — posture, skill, memory, hooks
python3 {skill_dir}/scripts/scan.py \
  --modules posture,skill,memory,hooks \
  --scan-dir . \
  --no-llm \
  --format markdown

# Full scan with LLM-enhanced analysis (uses ANTHROPIC_API_KEY automatically)
python3 {skill_dir}/scripts/scan.py \
  --modules posture,skill,memory,hooks,model \
  --scan-dir . \
  --format html \
  --output /tmp/deepsafe-report.html

# Scan only hooks (fastest — checks agent config files in current directory)
python3 {skill_dir}/scripts/scan.py \
  --modules hooks \
  --scan-dir . \
  --no-llm \
  --format markdown

# OpenClaw users (auto-reads gateway config)
python3 {skill_dir}/scripts/scan.py \
  --openclaw-root ~/.openclaw \
  --format html \
  --output /tmp/deepsafe-report.html

Replace `{skill_dir}` with the actual path to this directory.

After scanning

Present the total score and per-module breakdown
List CRITICAL and HIGH findings first with clear explanations
For each finding, explain what the risk is and how to fix it
Offer to help fix issues

Score interpretation

Score	Risk Level
85–100	LOW RISK
65–84	MEDIUM RISK
40–64	HIGH RISK
4–39	CRITICAL RISK

LLM API auto-detection

The scanner automatically uses `ANTHROPIC_API_KEY` if set (Claude Code users always have this). No manual configuration needed for:

LLM-enhanced skill semantic analysis
Model safety probes (persuasion, sandbagging, deception, hallucination)