drupal-security-audit

name	drupal-security-audit
description	Comprehensive security auditing workflow for Drupal sites. Covers user accounts, role permissions, module inventory, site configuration, text formats, views access, file exposure, content injection scanning, and integration with the security_review contrib module. Use when performing security assessments, hardening a Drupal site, or investigating potential vulnerabilities.
version	1.0.0
source	{"module":"ai_agents_security_review","agent":"security_reviewer","repository":"ai_agents_experimental_collection"}

Drupal Security Audit

Perform comprehensive, read-only security assessments of a Drupal site. Produce a clear, actionable report organized by severity. Never make changes to the site during an audit.

Audit Workflow

A complete security audit follows five phases. Each phase can also be run independently for targeted reviews.

Phase 1: Security Review Module

The security_review contrib module provides a suite of automated checks. When it is available:

1. List available checks to see every registered check with its ID, title, skip status, and last result.
2. Run all checks (or a subset by ID) to get a pass/fail/warn summary.
3. Get detailed findings for any check that returned FAIL or WARN, including remediation advice.

If the module is not installed, skip to Phase 2 -- all custom audits work independently.

Phase 2: User and Permission Audit

Audit user accounts and role permissions for common misconfigurations:

• UID 1 (super-admin): Verify the account is blocked or has a non-default username. An active uid 1 with the name "admin" is a predictable target.
• Privileged users: Identify every account holding an administrator role. Flag active admins that have never logged in and recently created accounts with admin roles.
• Recent accounts: Review accounts created in the last 30 days. New accounts with elevated roles warrant immediate investigation.
• Blocked accounts: Review the blocked-user list for patterns (e.g., compromised accounts that were disabled).
• Role permissions: Scan every role for dangerous permissions (administer modules, bypass node access, use PHP for settings, synchronize configuration, etc.). Permissions on the anonymous and authenticated roles are especially critical.

Dangerous Permissions Reference

Severity	Permissions
Critical	use PHP for settings, administer modules, administer software updates, administer site configuration, administer permissions, administer users, bypass node access, synchronize configuration, import configuration, administer filters
Warning	administer content types, administer taxonomy, administer views, administer blocks, access site reports, administer menu, administer nodes, use text format full_html, access all views, administer search

Any administer * permission on anonymous or authenticated roles is a critical finding.

Phase 3: Configuration and Module Audit

Review site-level settings and the module landscape:

• Site configuration: Check error-reporting level (should be hidden from visitors), trusted-host patterns (must be set), cron last-run time, private file path configuration, and PHP display_errors / expose_php settings.
• Input formats (text formats): Audit every text format for missing HTML sanitization, the PHP code filter, dangerous allowed tags, and which roles have access.
• Views access: Find enabled Views displays with no access plugin configured (access set to none). These are publicly reachable without restriction.
• Enabled modules: Flag development modules (devel, kint, webprofiler, field_ui, views_ui, dblog in production) and unsupported / dev-version modules.
• Security updates: Query Update Manager data for modules and core with known security advisories. Report current version vs. recommended version.
• Dangerous files: Scan the webroot for exposed sensitive files (.git/, .env, phpinfo.php, SQL dumps, backup archives, CHANGELOG.txt, INSTALL.txt).

Phase 4: Content Scanning

Scan content entities for potential stored XSS or PHP injection:

• Search text fields for patterns: <script, <?php, <iframe, javascript:, eval(.
• Default scope is nodes, but any fieldable entity type can be scanned.
• Limit the scan size (default 100 entities) to avoid performance impact.

Phase 5: Compile the Report

Organize all findings into a severity-ranked report:

Section	Contents
Executive Summary	Overall security posture with Critical / Warning / Info counts
Critical Issues	Must-fix immediately -- data exposure, code execution risks
Warnings	Should be addressed soon -- information disclosure, best-practice violations
Informational	Nice-to-have improvements
Recommendations	Prioritized action items with specific remediation steps

For every finding, explain why it is a security concern, not just what was found.

Tool Inventory

The audit relies on 12 dedicated read-only tools plus 2 module-management helpers:

#	Tool	Operation	Description	Key Parameters
1	list_security_checks	Read	Lists all security_review checks with ID, title, skip status, and last result	--
2	run_security_review	Read	Runs all or specific security_review checks; returns pass/fail/warn/info summary	checks: "all" or JSON array of check IDs
3	get_check_details	Read	Gets detailed findings and remediation help for a single check	check_id (required): plugin ID, e.g. "admin_user"
4	audit_users	Read	Audits user accounts: uid 1 status, privileged users, recent accounts, blocked accounts	audit_type: "all", "uid1", "privileged", "recent", "blocked"
5	audit_role_permissions	Read	Scans roles for dangerous permissions; flags anonymous/authenticated escalations	role: machine name or "all"
6	check_site_configuration	Read	Checks error reporting, trusted hosts, cron, private files, PHP settings	--
7	check_input_formats	Read	Audits text formats for missing HTML filtering, PHP code filter, dangerous tags	--
8	check_views_access	Read	Finds Views displays with no access control	--
9	audit_enabled_modules	Read	Flags development and risky modules in production	--
10	check_security_updates	Read	Reports modules/core with known security vulnerabilities via Update Manager	--
11	check_dangerous_files	Read	Scans webroot for .git/, .env, phpinfo.php, SQL dumps, backups, info files	--
12	scan_content_for_injections	Read	Scans entity text fields for <script, <?php, <iframe, javascript:, eval(	entity_type: defaults to "node"; limit: defaults to "100"
13	enable_module	Write	Enables a Drupal module (from ai_agents_module_manager)	module name
14	disable_module	Write	Disables a Drupal module (from ai_agents_module_manager)	module name

Tools 1-3 require the security_review contrib module to be enabled. Tools 4-12 work without any contrib dependencies. Tools 13-14 are used only for temporarily enabling/disabling security_review when needed.

Example Workflows

Full Site Audit

"Run a complete security audit of the site"

Execute all five phases in order. If security_review is not already enabled, enable it for the audit and disable it when finished. Compile a full severity-ranked report.

User Account Investigation

"Audit all user accounts for suspicious activity"

Use audit_users with audit_type="all" to check uid 1 status, list all privileged users, review recently created accounts, and inspect blocked accounts. Follow up with audit_role_permissions to check what each role can do.

Permission Hardening

"Check the permissions assigned to each role for dangerous access"

Run audit_role_permissions with role="all". Focus on anonymous and authenticated roles first. Any administer * permission on these roles is a critical finding. Provide specific remediation: which permissions to revoke and why.

Module Security Check

"Check if there are any security updates available"

Run check_security_updates to query Update Manager data. For each finding, report the project name, current version, and recommended security release. Also run audit_enabled_modules to flag dev-version or unsupported modules.

Configuration Review

"Check the site configuration for security issues"

Run check_site_configuration, check_input_formats, and check_views_access together. Look for error messages exposed to visitors, missing trusted-host patterns, text formats with the PHP code filter, and Views with no access restrictions.

Content Injection Scan

"Scan the content for potential XSS or injection attacks"

Run scan_content_for_injections with default parameters to scan up to 100 nodes. For larger sites, increase the limit or target specific entity types. Report any entities containing suspicious patterns with their entity ID and field name.

Webroot File Exposure

"Are there any dangerous files in the web root?"

Run check_dangerous_files to scan for .git/, .env, phpinfo.php, SQL dumps, backup archives, and Drupal info files that should not be publicly accessible. Each finding includes the file path and the security risk it poses.

Guidelines

• Read-only: Never modify site state during an audit (except temporarily enabling/disabling security_review when needed).
• Explain the risk: For every finding, explain the attack vector and potential impact, not just the misconfiguration.
• Provide remediation: Include specific, actionable steps to resolve each issue.
• Prioritize by risk: Rank findings by actual exploitability, not just severity labels.
• Continue on failure: If a tool fails or a module is missing, note it in the report and proceed with the remaining checks.
• Clean up: If security_review was enabled for the audit, disable it afterward to leave the site in its original state.