redveil

web vulnerability scanner. find vulns, validate safely, get a report you can actually send to a dev team.

$ pip install redveil
$ redveil scan https://target.example --scope scope.yaml
$ redveil list-checks

install

pip install redveil

or from source:

git clone https://github.com/FitzyPrjkt/Redveil
cd redveil
python3.12 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

requires python 3.12+.

quick start

# 1. write a scope file
cat > scope.yaml <<'EOF'
target:
  base_url: https://staging.example.com
scope:
  allowed_hosts:
    - staging.example.com
  allowed_paths:
    - /api/*
    - /account/*
limits:
  requests_per_second: 2
  max_requests: 500
authorization:
  active_testing: false
  acknowledged_safety_terms: false
profile: passive
EOF

# 2. scan
redveil scan https://staging.example.com --scope scope.yaml

# 3. results
ls reports/staging.example.com/
cat reports/staging.example.com/summary.md
open reports/staging.example.com/report.html

what you get

• 17 built-in checks — security headers, CORS, info disclosure, HTTP methods, open redirect indicators, source map exposure, XSS (canary reflection), SQLi (time-based), SSRF (OOB), command injection (time-based), path traversal (canary), BOLA/IDOR, BFLA, GraphQL, mass assignment, session/cookie config, subdomain discovery
• multi-format reports — markdown per finding, JSON for tooling, self-contained HTML
• strict scope enforcement — host + path allowlist, redirect chain validation, destructive path heuristic. plugins cannot bypass it
• multi-principal auth for BOLA testing — define Account A + Account B in scope, redveil compares what each can see
• evidence sanitization — JWTs, AWS keys, GitHub tokens, credit cards, cookies, emails all redacted before report
• local lab at tests/lab/ — a Flask app with 17 deliberately vulnerable endpoints for testing without hitting the internet

safety

redveil is a defensive tool. the active checks (XSS, SQLi, SSRF, command injection, path traversal) use bounded non-destructive payloads:

• XSS: alphanumeric canary strings. no <script>, no execution
• SQLi/command injection: time-based delay only (sleep 3). no data extraction
• SSRF: OOB callback to operator's own domain. no internal IP probing
• path traversal: unique canary filenames. no real file reads

runtime assertions in each check verify these constraints on every import. the test suite has explicit safety tests for every check.

you are responsible for authorization. redveil includes guards but they only matter if you actually have permission to test the target.

see SECURITY.md for the full safety model and how to report issues.

CLI

$ redveil scan <url> [--scope FILE] [--profile PROFILE] [--output DIR]
$ redveil check <plugin-id> <url>
$ redveil list-checks
$ redveil findings <report-dir>
$ redveil report <report-dir>

profiles: passive (default, observation only), low_impact (safe probes), active (requires explicit active_testing: true in scope).

writing checks

a check is a Check subclass:

from redveil.plugins.base import Check, CheckCategory, CheckMeta, ...

class MyCheck(Check):
    meta = CheckMeta(
        id="my-check",
        name="My Check",
        category=CheckCategory.HEADERS,
        safety_profile=SafetyProfile.PASSIVE,
    )
    async def discover(self, ctx): ...
    async def validate(self, ctx, candidate): ...
    async def collect_evidence(self, candidate): ...
    async def assess(self, candidate): ...

register in pyproject.toml:

[project.entry-points."redveil.checks"]
my-check = "my_pkg.checks:MyCheck"

see CONTRIBUTING.md for the full plugin spec.

files

• USER_GUIDE.md — installation, configuration, CLI reference, output interpretation
• CONTRIBUTING.md — how to add checks
• PUBLISH.md — how to publish a new release
• SECURITY.md — safety model, how to report issues
• CHANGELOG.md — release notes
• docs/architecture.md — internal design
• examples/ — scope files for common scenarios
• tests/lab/ — vulnerable Flask app for local testing

status

17 checks, ~1090 tests passing, 0 known safety violations. actively used against staging environments. the framework ships with curated, tested-safe payloads; destructive actions require per-action typed confirmation (no batch approval) and an explicit allow_destructive: true unlock in config.

what makes redveil different from sqlmap / nikto / burp scanner

see USER_GUIDE.md and docs/architecture.md for details.

license

MIT. see LICENSE.