orel-cybersecurity-check 0.1.3

oscan - a manual, authorized self-assessment scanner for web apps and sites (vulnerabilities + GDPR), with secure-by-default guidance.

oscan - orel-cybersecurity-check

Find the security holes AI-built apps ship with, in one command. Both a scanner and a Claude Code skill: audit any site, web app, or repo for vulnerabilities, leaked secrets, and GDPR gaps, or build new apps secure by default.

uvx --from orel-cybersecurity-check oscan https://your-site.example

A manual, authorized self-assessment scanner for the sites, web apps, and repos people increasingly build with AI. It catches the holes those builds ship with: SQL injection, secrets in git history, weak cookies/tokens, an exposed .git/.env, missing security headers, no rate-limiting, and it checks GDPR/privacy basics too.

Highlights

• One command, whole surface - TLS, security headers, cookies & tokens, secrets in git history, exposed .git/.env, injection, auth hardening, DoS-resilience, and GDPR.
• Scanner + Claude Code skill - audit an existing app, or build a new one secure by default.
• Actionable output - a 0-100 risk score, a letter grade, and plain-language fixes, plus JSON and Markdown reports.
• Safe by default - passive is observe-only; intrusive checks are gated behind --i-am-authorized.
• Local or remote - scan a live URL, a local repo, or a remote repo by its git URL.

Detection is deterministic (in code); an optional Claude layer only rewrites findings into plain language. The tool works fully offline without it.

This repo ships both pieces:

• oscan - the scanner (this README).
• orel_cybersecurity_check - a Claude Code skill (in skill/) that drives oscan and adds secure-by-default build guidance. See Claude Code skill below.

⚠️ Authorization & ethics

Use this only on assets you own or are explicitly authorized to test. It is for self-assessment, authorized pentesting, and learning.

• The default passive profile only observes (safe GET requests, no payloads).
• The intrusive active profile (injection markers, auth-lockout, a capped rate-limit burst) requires the --i-am-authorized flag, an interactive confirmation, and writes an audit log entry.
• There is no DDoS flooder and no password cracker. "DoS resilience" verifies rate-limiting/WAF exists (a hard-capped burst); "auth hardening" verifies account lockout exists (a few failed logins on your own account). The goal is to confirm defenses work, not to attack.
• All payloads are non-destructive detection markers - never DROP, never data exfiltration.

Install

Run it once, with no permanent install (needs uv):

uvx --from orel-cybersecurity-check oscan https://your-site.example

Install the oscan command for repeated use:

pipx install orel-cybersecurity-check     # isolated (recommended)
pip install orel-cybersecurity-check      # or plain pip

For the optional Claude enrichment layer, add the ai extra: pipx install "orel-cybersecurity-check[ai]".

From source, for development:

git clone https://github.com/orelsv/orel-cybersecurity-check.git
cd orel-cybersecurity-check
python3 -m venv .venv && source .venv/bin/activate
pip install -e .            # add ".[ai]" for the optional Claude layer

Optional external tools (auto-detected, graceful fallback if missing): gitleaks for stronger secret scanning.

Usage

# Passive scan of a live site (safe default)
oscan https://your-site.example

# Scan a local repo for secrets in the working tree AND git history
oscan --repo /path/to/your-repo

# Scan a remote repo by URL (cloned to a temp dir, then removed)
oscan --repo https://github.com/org/project

# Write machine + human reports
oscan https://your-site.example --json report.json --md report.md

# GDPR / privacy focus
oscan https://your-site.example --gdpr

# Active (intrusive, non-destructive) scan of YOUR OWN target
oscan https://your-site.example --profile active --i-am-authorized

# Verify login lockout exists (your own test account)
oscan https://your-site.example --profile active --i-am-authorized \
  --login-url https://your-site.example/login --login-user [e-mail]

Profiles

Profile	What it does	Gate
passive (default)	TLS, headers, cookies/tokens, exposed .git/.env, secrets, GDPR	none
standard	+ path enumeration, JWT analysis, WAF/rate-limit header check	none
active	+ SQLi/XSS/traversal/open-redirect markers, admin-endpoint probe, login-lockout, DoS-resilience burst	--i-am-authorized + confirm

What it checks

• Transport & TLS - HTTPS, TLS version, cert validity, HTTP→HTTPS redirect, HSTS
• Security headers - CSP, X-Content-Type-Options, clickjacking protection, Referrer-Policy, version disclosure
• Cookies & tokens - Secure / HttpOnly / SameSite flags, JWT in cookies, tokens in URLs
• Secrets & git - working tree and full git history (committed-then-deleted secrets), .gitignore coverage. Allowlist a known/fake secret with an inline # gitleaks:allow (or # pragma: allowlist secret) marker, or a repo .gitleaksignore.
• Exposed files - /.git, /.env, config/backup files, source maps
• CORS (standard) - arbitrary-origin reflection, wildcard-with-credentials
• API security (standard) - exposed OpenAPI/Swagger docs, GraphQL introspection, excessive data exposure (sensitive fields in JSON), verbose error/stack-trace disclosure
• Injection (active) - SQLi (error-based + time-based blind), NoSQL injection, reflected XSS, path traversal, open redirect, SSRF (cloud-metadata)
• Auth hardening (active) - JWT hygiene (alg=none, HMAC alg-confusion risk, empty signature, no-exp, sensitive claims), unauthenticated admin/debug endpoints, login rate-limiting/lockout
• DoS resilience - WAF/CDN/rate-limit presence + a capped burst to confirm throttling
• GDPR / privacy - consent before tracking, third-party trackers, Google Fonts, privacy policy, PII over HTTP

Output

Console summary with a Risk Score (0-100), a letter grade, and an estimated time-to-compromise, plus optional JSON (SIEM-friendly) and Markdown reports.

Optional Claude enrichment

If the anthropic package is installed and ANTHROPIC_API_KEY is set, each finding gets a one-sentence plain-language explanation. Detection is unchanged - Claude is only a presentation layer. Disable with --no-enrich; pick a model with OSCAN_ENRICH_MODEL (default claude-haiku-4-5).

Claude Code skill

The orel_cybersecurity_check skill lets Claude Code build apps securely by default and drive oscan to audit existing ones. Install it globally (one line):

mkdir -p ~/.claude/skills && curl -sL https://github.com/orelsv/orel-cybersecurity-check/archive/main.tar.gz | tar -xz --strip-components=2 -C ~/.claude/skills orel-cybersecurity-check-main/skill/orel_cybersecurity_check

Then restart your Claude Code session. The skill is at skill/orel_cybersecurity_check/SKILL.md. For a project-local install instead of global, replace ~/.claude/skills with .claude/skills in your project root.

Development

pip install -e ".[dev]"
ruff check . && ruff format --check .   # lint + formatting (CI gate)
pytest                                  # 51 tests, no network required

The test suite includes an end-to-end run against a deliberately vulnerable local server and a git repo with a committed-then-deleted secret. CI runs the same checks on Python 3.11-3.13. See CONTRIBUTING.md.

License

MIT