A fixed, visible rubric
Oracle rubric v1 starts at 100 and subtracts the fixed deduction attached to each scored finding. Total deductions are capped at 100, so the displayed score remains between 0 and 100.
ORACLE RUBRIC V1 · PUBLIC METHODOLOGY
A Discord audit score that shows its work.
Oracle’s score is a versioned product rubric derived from named configuration findings. It is not an industry benchmark, a guarantee of safety, or a substitute for reviewing coverage and evidence.
✓Read-only first✓Audit reads no messages✓Free during beta
THE CORE MODEL
Score, severity, coverage, and evidence stay separate
A single number cannot describe a Discord server responsibly. Oracle presents four related signals so owners can distinguish a scored configuration defect from an advisory observation or a section that was never available.
Assessed, partial, or unavailable
Every coverage section carries a status, limitations, and supporting evidence. Unavailable coverage does not lower the score, and must be read alongside it.
Prioritized without hiding advisories
Critical, high, medium, low, and informational findings are sorted by severity and deduction. New advisory rules may deliberately carry zero points.
EXPERT REVIEW AREAS
Five operating domains organize the consultant report
Safety & moderation
Guild safety settings, native AutoMod protection, rule exemptions, disabled-rule inventory, and age-restricted channel configuration.
Permissions & access
Role grants, hierarchy, bot manageability, role-by-channel perspectives, channel exposure, and duplicate permission profiles.
Structure & scale
Role, channel, and category headroom plus structural references that can become operational problems as the server grows.
Onboarding
Welcome-screen and native Community onboarding readiness, including Discord’s configuration requirements when evidence is readable.
Bots & integrations
Credential-free webhook inventory, connected applications, bot managed-role posture, and revoked or disabled integrations.
Unavailable behavioral coverage
Message content, member assignments, and activity are not treated as clean simply because Oracle intentionally does not collect them.
REMEDIATION POLICY
A finding is not permission to change the server
The audit engine never mutates Discord. It may attach a conservative remediation target only where the current product has an unambiguous safe operation.
Guild safety settings
Selected verification, notification, and explicit-content settings can enter a separate exact-preview and approval flow.
Oracle-owned AutoMod baselines
Oracle can preview and manage its own harmful-content and mention-spam baseline rules while preserving unrelated owner rules.
Roles, channels, and integrations
These findings remain owner guidance. Oracle does not mass-edit access, delete resources, or convert advice into hidden operations.
VERSIONING
Reports preserve the rubric and feature-set version
Each saved audit records its schema version, oracle-rubric-v1, and the feature-set version used for the run. This keeps a historical report interpretable when Oracle later adds rules or capabilities.
How to read a report responsibly
- Start with coverage: confirm the sections relevant to your decision were actually assessed.
- Review critical and high evidence before using the overall score as a summary.
- Separate safe-preview findings from recommendations that require owner judgment.
- Keep the result as a configuration baseline and rescan after meaningful changes.
For a concrete walkthrough, open the illustrative sample report. No real customer or guild data appears there.
FREE DURING BETA
Run the rubric against configuration Oracle can actually see.
The audit is private, read-only first, and included free during beta.
Run the free audit