| name | security |
| description | Expert guidance for writing secure Rails applications. Use when dealing with security, CSRF protection, XSS prevention, SQL injection, authentication, authorization, sanitize, html_safe, credentials, secrets, content security policy, session security, mass assignment, strong parameters, secure headers, file uploads, open redirects, or vulnerability remediation. Covers every major attack vector and the Rails-idiomatic defenses. |
| allowed-tools | Read, Grep, Glob, Write, Edit, Bash(bin/rails credentials:*), Bash(bin/rails secret), Bash(bundle audit), Bash(brakeman) |
Rails Security Expert
Write secure Rails code by default. Security mistakes are the most dangerous mistakes an agent can make — an XSS hole or SQL injection can compromise every user instantly. When in doubt, choose the safer path.
Philosophy
Core Principles:
- Secure by default — Rails has excellent built-in protections. Never disable them without explicit, documented justification
- Defense in depth — Layer protections. Don't rely on a single mechanism
- Permit-list over deny-list — Always prefer allowed lists over blocked lists for input validation, URL schemes, HTML tags, redirect targets
- Never trust user input — Every param, header, cookie, and URL segment is attacker-controlled until proven otherwise
- Fail closed — When authorization is ambiguous, deny access. When input is suspicious, reject it
- Minimize exposure — Log less, expose less, store less. Filter sensitive params, avoid leaking stack traces, encrypt at rest
When To Use This Skill
- Adding authentication or authorization logic
- Writing controllers that accept user input
- Rendering user-generated content in views
- Building API endpoints
- Working with file uploads or downloads
- Configuring sessions, cookies, or CSRF protection
- Managing secrets and credentials
- Setting HTTP security headers or CSP
- Reviewing code for security vulnerabilities
- Fixing reported CVEs or security issues
Instructions
Step 1: Audit the Existing Security Posture
Check what's already in place before changing anything — understanding the existing security posture prevents accidentally weakening it:
rg "protect_from_forgery" app/controllers/
rg "params\.permit\|params\.require" app/controllers/
rg "\.html_safe|raw\(|raw " app/views/ app/helpers/
rg "\.where\(\".*#\{" app/ --type ruby
rg "redirect_to.*params" app/controllers/
rg "content_security_policy" config/
rg "force_ssl" config/
cat config/initializers/filter_parameter_logging.rb
ls config/credentials* config/master.key 2>/dev/null
Step 2: CSRF Protection
Never disable CSRF protection. Rails enables it by default — it's your primary defense against cross-site request forgery.
class ApplicationController < ActionController::Base
protect_from_forgery with: :exception
end
How it works:
- Every non-GET request requires an authenticity token
- Include
<%= csrf_meta_tags %> in your layout <head> for Turbo/Ajax
- API controllers using token auth may skip CSRF — but only if they don't use cookie-based sessions (cookie sessions + no CSRF = attackers can forge requests)
- Never use
skip_forgery_protection on controllers that serve browser sessions
class Api::BaseController < ActionController::API
end
class PaymentsController < ApplicationController
skip_forgery_protection
end
Step 3: XSS Prevention
Rails auto-escapes all ERB output by default. This is your primary XSS defense — bypassing it means any user-controlled string can inject scripts into other users' browsers.
<%= user.name %>
<%= user.bio %>
<%= raw(content) %>
<%= content.html_safe %>
<%== content %>
<%= sanitize(user.bio, tags: %w[p br strong em a ul ol li], attributes: %w[href title]) %>
html_safe discipline:
- NEVER call
.html_safe on user input — this is an instant XSS hole
- NEVER call
.html_safe on strings built with interpolation of user data — the interpolated values bypass escaping
- Only use
.html_safe on string literals you fully control
- When building HTML in helpers, use
tag builder or content_tag — they handle escaping automatically
def greeting(user)
"<h1>Hello, #{user.name}</h1>".html_safe
end
def greeting(user)
tag.h1("Hello, #{user.name}")
end
def greeting(user)
"<h1>Hello, #{ERB::Util.html_escape(user.name)}</h1>".html_safe
end
sanitize() usage:
- Specify allowed tags and attributes explicitly — the default allowlist may be broader than you need
- For rich text, prefer Action Text which handles sanitization automatically
Step 4: SQL Injection Prevention
Never interpolate user input into SQL strings. Interpolation lets attackers inject arbitrary SQL — they can read, modify, or delete any data in your database.
User.where("name = '#{params[:name]}'")
User.where("email LIKE '%#{params[:q]}%'")
Order.where("status = '#{status}' AND user_id = #{user_id}")
User.where("name = ?", params[:name])
User.where("email LIKE ?", "%#{User.sanitize_sql_like(params[:q])}%")
Order.where(status: status, user_id: user_id)
User.where("name = :name AND role = :role", name: params[:name], role: params[:role])
User.where(name: params[:name])
LIKE queries need special handling:
User.where("name LIKE ?", "%#{params[:q]}%")
User.where("name LIKE ?", "%#{User.sanitize_sql_like(params[:q])}%")
Less obvious injection vectors — these methods also pass strings directly to SQL:
order() — never pass raw user input (attackers can inject subqueries)pluck(), select(), group(), having(), joins() with string argsfind_by_sql(), connection.execute()Arel.sql() — only use with trusted strings, never user input
Post.order(params[:sort])
allowed = %w[created_at title updated_at]
direction = %w[asc desc].include?(params[:dir]) ? params[:dir] : "asc"
column = allowed.include?(params[:sort]) ? params[:sort] : "created_at"
Post.order("#{column} #{direction}")
Step 5: Mass Assignment Protection
Use strong parameters for every model write. Raw params let attackers set any attribute — including admin flags, roles, and associations they shouldn't touch.
User.create(params[:user])
@user.update(params[:user])
User.new(params.permit!)
def user_params
params.require(:user).permit(:name, :email, :bio)
end
@user.update(user_params)
Strong params guidelines:
- Define a private
*_params method in every controller
- Only permit attributes the user should be able to set
- NEVER permit
:role, :admin, :is_admin, :verified or similar privilege attributes from user input — this is how privilege escalation happens
- Nested attributes need explicit permitting
def post_params
params.require(:post).permit(
:title, :body, :published,
tags: [],
images_attributes: [:id, :url, :caption, :_destroy]
)
end
Step 6: Authentication
Use the Rails 8+ authentication generator as your starting point:
bin/rails generate authentication
Key authentication practices:
- Use
has_secure_password — it handles bcrypt hashing correctly
- Use
authenticate_by — it's timing-safe, preventing attackers from enumerating valid accounts by measuring response time
- Call
reset_session after login to prevent session fixation (where an attacker pre-sets a session ID)
- Require current password for password/email changes
- Use generic error messages: "Invalid email or password" (not "User not found") — specific messages reveal which accounts exist
class SessionsController < ApplicationController
def create
if user = User.authenticate_by(
email_address: params[:email_address],
password: params[:password]
)
reset_session
start_new_session_for user
redirect_to root_path
else
redirect_to new_session_path, alert: "Invalid email or password."
end
end
end
Rate limiting login attempts:
class SessionsController < ApplicationController
rate_limit to: 10, within: 3.minutes, only: :create
end
Step 7: Authorization
Scope queries to the current user. Without scoping, any authenticated user can access any record just by changing the ID in the URL.
@project = Project.find(params[:id])
@project = Current.user.projects.find(params[:id])
@project = authorize Project.find(params[:id])
Authorization checklist:
- Scope every query to the authenticated user's permissions
- Check authorization on every action (before_action)
- Use
before_action :authenticate for all protected resources
- Prefer deny-by-default:
before_action in ApplicationController with selective skip_before_action is safer than opt-in only: (new actions are protected by default instead of accidentally exposed)
class ApplicationController < ActionController::Base
before_action :require_authentication
end
class PublicPagesController < ApplicationController
skip_before_action :require_authentication, only: [:home, :about]
end
Step 8: Session Security
config.force_ssl = true
Rails.application.config.session_store :cookie_store,
key: "_myapp_session",
secure: Rails.env.production?,
httponly: true,
same_site: :lax
Session guidelines:
- Store only IDs in the session — sensitive data in cookies is readable even when signed
- Call
reset_session after login (prevents session fixation)
- Implement session expiry (absolute + idle timeout)
force_ssl = true in production — without it, session cookies travel over plain HTTP and can be stolen on public networks
Step 9: Secrets and Credentials
Use Rails credentials or environment variables for secrets. Hardcoded secrets end up in version control, where anyone with repo access (or a leaked backup) gets your API keys.
EDITOR=vim bin/rails credentials:edit
EDITOR=vim bin/rails credentials:edit --environment production
Rails.application.credentials.secret_key_base
Rails.application.credentials.dig(:aws, :access_key_id)
Rails.application.credentials.some_api_key!
Key management:
- NEVER commit
config/master.key or config/credentials/*.key — these decrypt all your secrets
- Add them to
.gitignore (Rails does this by default, but verify)
- Use
ENV["RAILS_MASTER_KEY"] in production
- Avoid logging or exposing credentials in error messages
- Rotate secrets immediately if they may have been exposed
Step 10: Content Security Policy
Rails.application.config.content_security_policy do |policy|
policy.default_src :self, :https
policy.font_src :self, :https, :data
policy.img_src :self, :https, :data
policy.object_src :none
policy.script_src :self, :https
policy.style_src :self, :https
policy.connect_src :self, :https
policy.frame_ancestors :none
policy.report_uri "/csp-violation-report-endpoint"
end
Rails.application.config.content_security_policy_nonce_generator =
->(request) { SecureRandom.base64(16) }
Rails.application.config.content_security_policy_nonce_directives = %w[script-src]
Layout must include:
<head>
<%= csrf_meta_tags %>
<%= csp_meta_tag %>
<%= javascript_include_tag "application", nonce: true %>
<%= stylesheet_link_tag "application", nonce: true %>
</head>
Step 11: Logging Security
Rails.application.config.filter_parameters += [
:passw, :secret, :token, :_key, :crypt, :salt, :certificate, :otp,
:ssn, :credit_card, :card_number, :cvv, :authorization
]
Avoid logging (these show up in log files, error trackers, and potentially third-party services):
- Passwords, tokens, API keys, secrets
- Credit card numbers, SSNs, PII
- Full request bodies that may contain sensitive data
Step 12: Redirect Security
redirect_to params[:return_to]
redirect_to safe_redirect_path(params[:return_to])
def safe_redirect_path(path)
uri = URI.parse(path.to_s)
if uri.relative? && uri.path.start_with?("/")
uri.path
else
root_path
end
rescue URI::InvalidURIError
root_path
end
redirect_to url_from(params[:return_to]) || root_path
Step 13: File Upload Security
class Document < ApplicationRecord
has_one_attached :file
validates :file, content_type: %w[application/pdf image/png image/jpeg],
size: { less_than: 10.megabytes }
end
File upload guidelines:
- Validate content type AND file extension — attackers can spoof one but rarely both
- Set maximum file size limits to prevent denial-of-service via large uploads
- Store uploads outside the web root (Active Storage does this automatically)
- Never execute uploaded files
- Process images/media asynchronously — image processing libraries can have vulnerabilities
- Scan for malware in sensitive applications
Quick Reference
Dangerous Patterns to Grep For
Regex Gotcha
validates :url, format: { with: /^https?:\/\/.+$/ }
validates :url, format: { with: /\Ahttps?:\/\/.+\z/ }
Anti-Patterns — Each of These Is a Vulnerability
html_safe on user input — Instant XSS. Attackers inject scripts that run in other users' browsers- String interpolation in SQL — SQL injection lets attackers read/modify/delete your entire database
params.permit! — Permits everything, letting attackers set admin flags, roles, or any attributeskip_forgery_protection on browser controllers — Lets malicious sites submit forms on behalf of your usersredirect_to params[:url] — Open redirect. Attackers craft links that look like your site but redirect to phishing pages- Hardcoded secrets in source — Anyone with repo access (or a leaked backup) gets your API keys
Kernel#open with user input — open("| rm -rf /") executes shell commands. Use File.open, URI.open, or IO.open- Disabling
force_ssl in production — Session cookies travel in plaintext, stealable on any shared network
- Logging sensitive data — Logs end up in error trackers, log aggregators, and developer machines
- Trusting client-side validation — Attackers bypass JavaScript trivially; server-side validation is the real gate
Security Audit Checklist
Before shipping any feature, verify:
