Security & data protection¶

The RAGflow Moodle suite is built to Moodle's secure-coding standards and reviewed continuously. This page is a transparency summary of the safeguards that are in place and the checks we run before every release. It is intentionally written as a checklist of what we verify and protect against — it does not contain any exploit-level detail.

How we verify¶

Every plugin in the suite is checked, on each change, by:

moodle-plugin-ci — the same pipeline the Moodle plugins directory uses: phpcs (Moodle coding standard), phpmd, phpdoc, validate, upgrade savepoints, Mustache, and eslint/stylelint, running on Moodle 5.0 and 5.2 × PostgreSQL and MariaDB.
Automated tests — PHPUnit unit tests and Behat acceptance tests ship with every plugin and run in CI.
The Moodle plugin submission guideline checker — run per plugin (0 blocking findings).
Periodic manual security audits against the Moodle-specific checklist below.

What we check and the safeguards in place¶

Access control & permissions¶

Every page and endpoint requires an authenticated session and an explicit capability check; site-wide administrative functions additionally require site-administrator rights.
Capabilities are declared with appropriate risk flags, role archetypes and context levels, following the principle of least privilege.
Web-service functions validate the calling context and enforce their capability before doing any work.

Input & output handling¶

All request input is read through Moodle's typed parameter API (PARAM_*); the code uses no raw PHP superglobals.
All output is HTML-escaped — via Moodle's s()/format_string() helpers and auto-escaping templates — and AI answers are sanitised server-side (scripts/dangerous markup removed) everywhere they are displayed.
Database access is fully parameterised; no SQL is built by string concatenation.

Cross-site request forgery (CSRF)¶

All state-changing actions run through Moodle's form and web-service framework, which enforces session-key (sesskey) tokens; any custom endpoint additionally requires a valid session key and capability.

File handling & downloads¶

Uploaded files pass Moodle's size limits, antivirus scanning and filename sanitisation.
Document downloads use short-lived, per-user signed tokens (verified with constant-time comparison) or context-authorised links; content types are restricted and served as attachments with nosniff; a user can only reach documents that belong to the knowledge base their instance is actually configured for.

Secrets & the external service¶

The RAGflow API key is stored server-side only. It is never sent to the browser and never written to any log — the optional troubleshooting log records request/response bodies only, never credentials.
Outbound calls go through Moodle's HTTP client to the administrator-configured RAGflow endpoint.

Safe deserialisation & code execution¶

Stored configuration is deserialised with a strict allow-list (plain data objects only); no untrusted input is ever deserialised.
There is no dynamic code execution (no eval, no shell calls) — verified automatically.

Abuse protection¶

Chat requests are rate-limited per user.

Every plugin implements Moodle's Privacy API. Locally stored data is exportable and deletable per user; data transmitted to the external RAGflow service (prompts, uploaded documents and their provenance) is declared in Moodle's privacy registry; and user deletion removes local records and forgets the user's RAGflow memory.
The usage dashboard offers an anonymisation option and an automatic retention/purge task; its optional raw request/response capture is off by default and admin-only.

AI-specific handling¶

Model output is treated as untrusted: sanitised before storage or display and never executed. Source citations link only to validated URLs.

Recommendations for administrators¶

The Helpdesk chat is available to authenticated users by default — review your site's self-registration settings so the audience matches your intent.
Enable the dashboard's raw API-call capture only while troubleshooting, and turn it off afterwards; anonymisation applies to newly captured rows.
Configure the RAGflow endpoint over HTTPS to a trusted host.

Reporting a security concern¶

If you believe you have found a security issue in any of these plugins, please report it privately to info@ragcon.ai rather than opening a public issue, and do not include exploit details in public channels. We will acknowledge and address verified reports promptly.

Note: This documentation was created with the help of AI. Spotted an error, an omission, or something unclear? Please report it in the ragcon-docs GitHub repository so we can fix it.