Editor AI WP Registry grade D. High-severity findings · 6 findingsWP RegistryHigh-severity findings · 6 findingsOpen the reportD

High-severity findings.

Audited release

1.0

Findings

6

Worst severity

high

Content hash

27bede83977912642b041f05… Copy

Findings

• high Unauthenticated POST /editor-ai/v1/playground proxies attacker payloads to Athena endpoint

  missing_permission_callback

  Rest.php registers /playground with no permission argument, so UseRest defaults to permission_callback __return_true. Playground::restCallback accepts JSON task/input and HttpClient::send posts it to get_option('editorai_athena_endpoint') with sslverify filtered to false. Unauthenticated users can burn AI credits, inflate traffic, and inject arbitrary prompt/model fields into the site's paid backend.

  src/Rest.php:22-22

  Recommendation

  Pass permission: 'edit_posts' or 'manage_options' to addRestRoute for /playground. Reject unauthenticated requests in Playground::restCallback as defense in depth.

• high Unauthenticated GET /editor-ai/v1/agents/test executes DictionaryAgent chat

  missing_permission_callback

  The /agents/test route is registered without a permission argument (defaults to __return_true) and immediately runs DictionaryAgent::make()->chat(new UserMessage(...)). Any anonymous visitor can trigger live LLM provider calls and receive serialized agent output/errors.

  src/Rest.php:27-53

  Recommendation

  Remove the test route from production builds, or require manage_options/edit_posts. Never leave provider-backed test endpoints publicly reachable.

• high Unauthenticated POST /editor-ai/v1/playground proxies attacker payloads to Athena endpoint

  missing_permission_callback

  Rest.php registers /playground with no permission argument, so UseRest defaults to permission_callback __return_true. Playground::restCallback accepts JSON task/input and HttpClient::send posts it to get_option('editorai_athena_endpoint'). Unauthenticated users can burn AI credits, inflate traffic, and inject arbitrary prompt/model fields into the site's paid backend.

  src/Rest.php:22-22

  Recommendation

  Pass permission: 'edit_posts' or 'manage_options' to addRestRoute for /playground. Also reject unauthenticated requests inside Playground::restCallback.

• high Unauthenticated GET /editor-ai/v1/agents/test executes DictionaryAgent chat

  missing_permission_callback

  The /agents/test route is registered without a permission argument (defaults to __return_true) and immediately runs DictionaryAgent::make()->chat(new UserMessage(...)). Any anonymous visitor can trigger live LLM provider calls and receive serialized agent output/errors.

  src/Rest.php:27-53

  Recommendation

  Remove the test route from production builds, or require manage_options/edit_posts.

• low Outbound AI HTTP client disables SSL verification by default

  insecure_config

  HttpClient::send sets sslverify via filter editorai_client_http_sslverify with default false, weakening TLS for requests that carry site AI traffic.

  src/Shared/Clients/Http/HttpClient.php:15-18

  Recommendation

  Default sslverify to true; only allow false behind an explicit admin opt-in.

• low Outbound AI HTTP client disables SSL verification by default

  insecure_config

  HttpClient::send sets sslverify via filter editorai_client_http_sslverify with default false, weakening TLS for requests that carry site AI traffic to the Athena endpoint.

  src/Shared/Clients/Http/HttpClient.php:15-18

  Recommendation

  Default sslverify to true; only allow false behind an explicit admin opt-in.

WP Registry hashes the installable build and reports on that exact bytes-for-bytes copy. Embargoed findings are withheld until they are disclosed, so a clean verdict means nothing public is outstanding. WP Manifest does not audit code itself.