SudoWP Radar vs VideoWhisper Security Audit

WordPress 7.0 introduced a new AI attack surface. Every plugin that registers an ability on your site declares a structured entry point for AI agents and MCP tools. SudoWP Radar audits that surface at runtime, flagging misconfigurations before they become incidents. It sits between reactive CVE scanners (which wait for a vulnerability to be disclosed) and developer-side static analysis tools (which run before deployment). Radar audits what is actually registered and executing on your live site, right now. What it audits Core ability rules (WP 6.9+) Open and weak permissions — abilities with no permission_callback, or one that passes any authenticated user through regardless of role. Missing or loose input schemas — abilities that accept unconstrained string inputs on fields like path, file, url, redirect, or slug. Common injection vector for path traversal and SSRF. REST overexposure — abilities marked show_in_rest with no or open permission control, reachable by unauthenticated callers. Orphaned callbacks — execute_callbacks referencing functions no longer loaded, typically left behind by deactivated plugins. Namespace collisions — duplicate ability names where the last registration silently overwrites the first, potentially downgrading the permission model. AI agent rules (WP 7.0+) AI prompt filter bypass (HIGH) — a plugin has disabled the sitewide AI prompt prevention gate. Any AI agent connected to your site bypasses this control. AI REST overexposure (CRITICAL/HIGH) — REST endpoints that invoke the AI client with no or weak permission checks. Directly exploitable by unauthenticated callers. AI missing version gate (MEDIUM) — plugins calling the WP 7.0 AI client without a compatibility check, causing fatal errors on sites not yet running 7.0. Hosting-injected ability (HIGH) — an ability registered by a plugin auto-installed by your hosting provider, without explicit site administrator consent, with REST exposure enabled. Requires premium vendor slug list via the SudoWP dataset. Connector key in database (HIGH) — an AI provider API key (OpenAI, Anthropic, Google, or similar) is stored as plaintext in your WordPress database via the WP 7.0 Connectors API. Any SQL injection or object cache exposure on your site leaks this key directly. The fix is to define it as an environment variable or PHP constant instead. Why this matters now WordPress 7.0 ships with native AI agent integration. Plugins can now register abilities that AI agents call directly, expose AI endpoints over REST, and connect to external AI providers via the Connectors API. Each of these is a new attack surface that existing security scanners do not cover — they match known CVEs, they do not audit AI agent architecture. Some hosting providers have begun auto-installing AI agent plugins on customer sites without explicit consent. If one of those plugins registers abilities with REST exposure, or stores an AI provider key in your database, Radar flags it. How it works Radar reads the live abilities registry after all plugins and themes have loaded. It applies its rule engine to each registered ability and returns a findings report with severity ratings (CRITICAL, HIGH, MEDIUM, LOW) and specific remediation guidance per finding. A risk score from 0-100 summarises the overall exposure. The audit runs on demand. It does not affect front-end performance. Security model Requires the radar_run_audit capability (administrators by default). All requests are nonce-gated. No public-facing endpoints. Findings are stored in user meta, not global options. Rate-limited to one audit per 30 seconds per user. Free vs premium The free plugin is a fully functional standalone auditor. An optional premium add-on (SudoWP Pro) extends it with vulnerability dataset matching (CVE references, CVSS scores, patch guidance), the hosting-injected vendor slug list, scheduled audits with email alerts, multi-site dashboard aggregation, and report export. None of the premium features are required to run the core audit. External Services When an API key is configured, SudoWP Radar connects to the SudoWP vulnerability dataset API (api.sudowp.com) to retrieve patch availability information for registered WordPress abilities. No data is transmitted without an API key being explicitly entered by the site administrator. When no key is present, the plugin makes zero external network requests. Data sent to the API: the ability name being looked up and your API key. No personal data, no site URL, no user data is transmitted. API key registration: https://sudowp.com/get-api-key/ Terms of service: https://sudowp.com/tos/ Privacy policy: https://sudowp.com/privacy-policy/

VideoWhisper Security Audit creates WordPress site health, exposure, vulnerability, integrity, readiness, and performance reports for site administrators. The plugin is designed to help administrators review site activity and configuration with AI agents or by using the built-in admin report. The free plugin is read-only. It reports findings and does not perform cleanup, quarantine, updates, file changes, role changes, or other remediation actions. Plugin homepage: https://promptaur.com/wordpress/security-audit/ Main features: Admin Scan tab with queued scan processing, live progress, per-task status, resume/retry controls, scan date, report summary, and report findings. Configurable scan execution mode: queued scans by default, or legacy synchronous scans for small/basic sites. Importance-based scan modes: full, critical/high/medium, and critical-only. Report filters for issues only or all check results, including passed informational checks when available. JSON and plain-text Markdown report output. Optional token-protected REST report endpoint. Optional token-protected MCP endpoint for read-only AI-agent reports. Optional WordPress 6.9+ Abilities for admin-only read-only security overview, vulnerability, exposure, integrity, readiness, performance risk, and Markdown audit reports. Optional exposure of those abilities through the official WordPress MCP Adapter when installed separately. Read-only integration with VideoWhisper AI Site Manager – Using ChatGPT Claude Codex when both plugins are active, including report reads and agent-friendly queued scan tools. Separate REST and MCP enable/disable settings. Generated local tokens with rotation controls and last-used metadata. REST/MCP per-minute rate limiting and optional exact IP allowlist. Admin scan cooldown, hourly scan limit, and separate agent scan cooldown. Category toggles for security, integrity, performance, readiness, commerce, community, and backup checks. Redacted AI report defaults, with optional exact version and path disclosure for MCP reports. Optional WPVulnerability API lookups for installed plugin vulnerability data. Disclaimers in the admin report and Agents tab about report limits, sensitive information, and third-party AI analysis. Queued scans Queued scanning is the default admin interface. It splits inventory, exposure, vulnerability, integrity, performance, and readiness checks into small AJAX-polled batches. This avoids browser, proxy, and PHP timeout issues on larger sites while showing progress and recent queue events. Failed tasks can be retried, and unfinished jobs can be resumed from the Scan tab. Administrators that prefer the original one-request scan flow can switch Scan execution to “Synchronous legacy scan” in Settings. Local checks Security Audit currently checks local WordPress signals including: Plugin and theme updates. Inactive plugins. Administrator account count. Expected WordPress database tables. Administrator role capabilities. Upload directory writability. Debug log file presence. Git metadata in the web root. XML-RPC availability. Common homepage security headers. Autoloaded option size. Expired transient count. WP-Cron disabled state. Permalink structure. Search engine visibility setting. Privacy Policy page presence. Basic WooCommerce page readiness when WooCommerce is active. Optional WPVulnerability plugin vulnerability lookups. Agent and API reports REST and MCP endpoints are disabled by default. When enabled, Security Audit automatically generates local tokens. Anyone with a valid token can read the selected report until the token is rotated, so treat tokens as sensitive secrets. The REST report endpoint supports: key: generated REST token. A bearer token can also be used. mode: full, important, critical, or changed. report: issues or all. format: omit for JSON, or use markdown for plain Markdown output. The MCP endpoint supports read-only tools for security summary, vulnerability, exposure, integrity, performance risk, readiness, and Markdown audit reports. Tool arguments include mode and report. When VideoWhisper AI Site Manager is active, Security Audit also registers agent tools for latest summary/report reads, synchronous scan refreshes for small sites, scan task discovery, queued scan creation, queue batch processing, queue status polling, and retrying failed scan tasks. On WordPress 6.9+, administrators can also enable WordPress Abilities. These abilities are admin-only, read-only, and permission-checked with manage_options by default. They can optionally be marked public for the official WordPress MCP Adapter, which must be installed separately. Existing REST and MCP endpoints remain available and are not replaced. Some sites use an OAuth, firewall, or bearer-auth layer that consumes Authorization: Bearer before the Security Audit route receives the request. If standalone Codex MCP bearer setup returns oauth_token_invalid or a similar upstream bearer-auth error, use the MCP path-token URL shown on the Agents tab instead of the bearer URL. Endpoint protection controls include: REST/MCP requests per minute per IP. Separate agent scan cooldown. Optional exact IPv4/IPv6 allowlist. Token rotation. Developer hooks Security Audit exposes generic hooks that any plugin can use to extend scan data, reports, admin UI, and Site Manager integration: vwsa_latest_snapshot filters the latest stored snapshot before reports use it. vwsa_scan_settings filters normalized scan settings for a scan context. vwsa_scan_inventory filters inventory data before findings are finalized. vwsa_scan_findings filters findings before snapshot lifecycle metadata is saved. vwsa_scan_queue_tasks filters task definitions shown and queued by the Scan tab. vwsa_scan_queue_process_task processes extension-owned queue tasks without coupling the extension to Free internals. vwsa_snapshot_before_save filters the completed snapshot before it is stored. vwsa_snapshot_saved fires after a snapshot is stored. vwsa_report filters built report output. vwsa_site_manager_integration filters the Site Manager integration definition. vwsa_admin_tabs filters admin tabs. vwsa_admin_render_tab lets extensions render custom admin tabs. vwsa_admin_scan_panel_after renders generic content after the scan panel. vwsa_admin_findings_before renders content before findings. vwsa_admin_finding_card_start renders content inside a finding card. vwsa_admin_finding_actions filters per-finding action links. vwsa_admin_findings_after renders content after findings. Important limitations and disclaimers Security Audit reports are informational only. Findings and AI-ready reports may be incomplete, inaccurate, outdated, or unsuitable for a specific site or legal situation. Security Audit does not provide legal advice, compliance certification, professional security advice, malware cleanup, incident response, or a guarantee that a site is secure. Administrators should verify findings and consult an experienced security, technical, or legal provider before making important changes. REST and MCP reports may expose sensitive operational information, including site configuration, component versions, possible vulnerabilities, paths, and other details. Enable agent endpoints only when you understand where the data will be sent and who can access the token. Third-party AI agents may produce incomplete, incorrect, unsafe, or unsuitable recommendations. Review all recommendations before acting and do not perform destructive changes without backups and appropriate professional review. This plugin is not a firewall, malware cleaner, legal compliance tool, vulnerability scanner guarantee, or replacement for backups, security monitoring, dedicated scanners, or experienced administrators. External services By default, Security Audit does not call external vulnerability services. If the administrator enables WPVulnerability lookups, the plugin sends installed plugin slugs to the public WPVulnerability API at https://www.wpvulnerability.net/ to retrieve vulnerability data. No API key is required for normal component lookups. Responses are cached locally. See: https://www.wpvulnerability.com/ https://docs.wpvulnerability.com/ https://www.wpvulnerability.com/privacy/ https://www.wpvulnerability.com/license/ During scans, Security Audit may also make a local HTTP HEAD request to the site’s own homepage URL to inspect response headers. This request is sent to the configured site URL, not to a third-party vulnerability service.