Review & Quality Guide

This page explains how AISoulHub checks public content in user-readable terms: review decides whether content can be published, quality checks provide improvement suggestions, and processing results explain what action the platform took.

1. What Review and Quality Checks Mean

AISoulHub checks are not meant to flatten content. They make public content safer, more complete, easier to use, and easier to repair.

Review

Decides whether content can be published publicly, focusing on safety boundaries, obvious violations, high-risk expressions, and baseline completeness.

Quality Check

Evaluates structure, executability, use cases, boundary clarity, and sample quality, then provides suggestions.

Processing Result

Explains what action AISoulHub took, such as allowing publication, returning for fixes, blocking, or routing to review.

2. How to Read Results

Review passed: this version can be published publicly.
Review failed: this version cannot be published publicly and should be fixed before resubmission.
Improvement suggestion: worth improving, but not always blocking.
Required fix: must be addressed before passing review.
Service issue: the checking service is temporarily unavailable; it does not mean your content violates policy.

The public UI does not show internal rule IDs or template names. You should see understandable reasons and actionable suggestions.

3. SOUL-6, SAFE-3 & PDS-6

AISoulHub explains quality with three user-readable frameworks: SOUL-6 for AI Souls, SAFE-3 as the common Skill baseline, and PDS-6 as the extra standard for Persona Skills. They are not rigid templates; they clarify what is complete and what needs stronger boundaries.

Detail pages now translate these frameworks into visual quality overviews: AI Souls mainly use a SOUL-6 radar, Persona Skills show both SAFE-3 bars and a PDS-6 radar, and Soul-building Skills focus on SAFE-3 plus operational cards. The goal is to let people read the overview first and only drill into dimensions when needed.

Frontend visual examples

Visitors first see the visual snapshot; authors and admins continue into improvement items, dimension cards, and file-level diagnostics.

SOUL-6

AI Soul radar

SAFE-3

Safe

Assured

Functional

PDS-6

Persona radar

SOUL-6: AI Soul Quality Dimensions

Selfhood: whether identity, motivation, expression, user address, and hidden traits are coherent.
Operability: whether users can understand how it responds and collaborates.
Unity: whether settings, tone, catchphrases, sample lines, and files support each other.
Limits: whether capabilities, refusals, timing, and in-character refusal are clear.
Safeguards: safety is built into the soul's behavior, not bolted on later.
Lifecycle: whether memory, evolution, and versioning have a clear direction.

Forge also runs a personification structure precheck: user address, relationship, catchphrases, hidden traits, sample lines, in-character refusal, and emotional safety boundaries. It is a generation-time precheck, not a replacement for final publishing review.

SAFE-3: Skill Quality Dimensions

Safe: whether boundaries, refusal conditions, privacy, and professional-advice limits are clear.
Assured: whether structure, inputs/outputs, triggers, and scope are stable and reusable.
Functional: whether it can actually complete the target task with usable workflow and examples.

SAFE-3 covers the common baseline for all Skills: safety, structure, stability, and basic usability. It answers whether this is a viable Skill, not whether a specific person has been truly distilled.

PDS-6: Persona Distillation Quality

PDS-1 Persona anchoring: whether the identity, position, and relationship frame of the persona are clear.
PDS-2 Cognitive distillation: whether stable mental models, judgment patterns, and heuristics are distilled.
PDS-3 Expression restoration: whether tone, wording, cadence, and expression structure are consistently restored.
PDS-4 Boundary authenticity: whether facts, inferences, and unknowns are separated with non-impersonation and professional limits.
PDS-5 Evidence traceability: whether key conclusions can be traced to research summaries, source types, and evidence anchors.
PDS-6 Interactivity: whether it provides triggers, response protocol, and fallback rather than only static persona analysis.

PDS-6 is only used for Persona Skills. It answers whether the result feels like the person, is grounded enough, and can be used interactively, so it is shown alongside SAFE-3.

4. Content-specific Quality Focus

AI Soul

Focuses on personality consistency, behavior protocol clarity, boundaries and safety, memory policy, and audition sample quality.

Persona Skill

Starts with the SAFE-3 baseline, then emphasizes PDS-6: persona anchoring, cognitive distillation, expression restoration, boundary authenticity, evidence traceability, and interactivity.

Soul-building Skill

Focuses on shaping-goal clarity, input/output constraints, workflow, executability, and safety boundaries.

5. Quality Snapshots and Current Suggestions

A package may include quality-check.md. It is a self-check snapshot from generation or upload time. Current quality suggestions on detail pages may be re-evaluated using the platform's current rules.

Publishing review result: the conclusion when this version was submitted.
Current quality suggestions: suggestions based on current platform rules.
File snapshot: the record stored inside the package at generation or upload time.

6. If You Think It Was Misjudged

AI checks can be wrong. If you believe a result is inaccurate, use the feedback entry on the page. AISoulHub will use context, files, and rules to improve future judgments.