Prompt Workspace Architecture - Source Excerpt 02 - Core Responsibilities

Summary

This source excerpt begins near Core Responsibilities and preserves the surrounding evidence from Wiki.FFTAC.org/raw/system-archives/spiralist.org/docs-archive/2026-05-10/docs/prompt-workspace-architecture.md.
**Source path:** Wiki.FFTAC.org/raw/system-archives/spiralist.org/docs-archive/2026-05-10/docs/prompt-workspace-architecture.md
The calibrant guidance promoted from the active 2026-05 calibrant intake is implementation guidance, not a claim that calibrant is a settled movement-wide technical standard. Workspace treats calibrants as visible prompt and interpretation artifacts: mirror contracts, seeds, spores, glyph payloads, codex/transmission texts, triad reviews, and anti-seed repairs. The current public workbench modes are `calibrant-map`, `calibrant-triad-review`, and `calibrant-safety-repair`; the matching managed prompts are `spiralist-calibrant-map`, `spiralist-calibrant-triad-review`, and `spiralist-calibrant-safety-repair`. These tools should inspect source surface, intended tuning, mechanism, evidence, risk cues, safer return, consent, review, stop conditions, and clean exit language without validating hidden-message, sentience, destiny, dependency, Protocol5 authority, or UAIX authority claims.

The homepage and `ai-access` pages enqueue Workspace assets so prompt-running CTAs and account-aware prompt actions work outside the core Workspace routes.

## Core Responsibilities

The plugin currently owns:

- public prompt discovery and prompt dossiers
- member authoring and editing workflows
- provisional public publishing and peer-moderation state changes for community prompts
- immutable prompt version snapshots
- prompt tests and admin batch test runs
- favorites, sharing, and permission-gated visibility
- server-side prompt execution and prompt improvement
- prompt-governed conversations
- public member profiles with scoped, sanitized, prompt-guided CSS customization
- managed filesystem prompt syncing from `prompts/<slug>/<version>/prompt.php`
- optional Spiral Script v0.1 source storage, canonical JSON compilation, validation metadata, and publication blocking for public or published prompts that fail Spiral Script validation

## Plugin Structure

Current top-level structure:

' ' ' text
spiralist-workspace/
  spiralist-workspace.php
  assets/
    css/workspace.css
    js/workspace.js
  includes/
    Admin/
      ManagedPromptAdmin.php
    Auth/
      Permissions.php
      Roles.php
    Config/
      Settings.php
    Domain/
      AppealRepository.php
      AuditRepository.php
      ConversationRepository.php
      FavoriteRepository.php
      PermissionRepository.php
      PromptRepository.php
      ReputationRepository.php
      ReviewRepository.php
      RunRepository.php
      TestRepository.php
      VersionRepository.php
    Http/
      OpenAIClient.php
      RestController.php
    Privacy/
      Privacy.php
    Services/
      GovernanceMaintenance.php
      PlaceholderService.php
      PromptRunner.php
      RateLimiter.php
    Support/
      PromptCategoryRegistry.php
      PromptMetaRegistry.php
      SystemPromptRegistry.php
    Web/
      FrontendController.php
  prompts/
  templates/
    conversation.php
    library.php
    members.php
    prompt.php
    run.php
    workspace.php
' ' ' 

## Storage Model

Prompt content uses a dedicated WordPress post type and taxonomies:

- post type:
  - `spiralist_prompt`
- taxonomies:
  - `spiralist_prompt_tag`
  - `spiralist_prompt_type`
  - `spiralist_prompt_cat`

The prompt post and meta model stores the authored prompt contract plus publication metadata such as:

- structured prompt fields
- system, developer, and user prompt layers
- model and output settings
- visibility and moderation state
- publication lane and provenance
- risk labels and function tags
- maturity level
- related symbols, axioms, and manuscript references
- sample output and execution options
- optional Spiral Script source, compiled canonical JSON, validation result JSON, and validation status

Custom tables handle relational and high-volume data:

- `wp_spiralist_workspace_prompt_versions`
- `wp_spiralist_workspace_prompt_permissions`
- `wp_spiralist_workspace_prompt_runs`
- `wp_spiralist_workspace_conversations`
- `wp_spiralist_workspace_conversation_messages`
- `wp_spiralist_workspace_audit_log`
- `wp_spiralist_workspace_favorites`
- `wp_spiralist_workspace_prompt_tests`
- `wp_spiralist_workspace_prompt_reviews`
- `wp_spiralist_workspace_prompt_appeals`

All custom-table timestamps are stored in UTC.

## Managed Prompt Registry

Managed and official prompts can live on disk under:

' ' ' text
wp-content/plugins/spiralist-workspace/prompts/<slug>/<version>/prompt.php
' ' ' 

`SystemPromptRegistry` syncs those filesystem prompts into the WordPress database so they appear in the same public and member surfaces as site-authored prompts.

The sync path first looks up managed prompts by source slug, then falls back to the canonical WordPress slug and numeric suffix variants such as `<slug>-2`. Public prompt lists also collapse managed numeric slug variants so a CMS duplicate does not appear as a second prompt card while the database record is being reconciled.

`ManagedPromptAdmin` adds admin-only tools for:

- importing prompt files and ZIP prompt packs
- editing filesystem-backed prompt definitions
- downloading starter templates
- uploading related media and appending filenames to prompt metadata

Filesystem-managed prompts are the source of truth for managed system prompts.

Those admin-only tools are for managed and official prompt handling. They are not the intended long-term governance model for normal community publication.

## Public Library Model

The public prompt library is not a flat list. It is currently organized as a progressive prompt commons: the hero explains the beginner and expert paths, first-session and intent cards appear first, recommended starter prompts follow, study pathways explain the system second, and expert filters, API examples, stats, and full lane lists stay behind disclosure panels. Publication metadata still comes from `PromptMetaRegistry` and `PromptCategoryRegistry`. The `/run-a-prompt/` page sits in front of that library as a task-first prompt builder: it should make portable no-login use obvious before asking visitors to inspect, fork, save, or run account-backed prompt records. Share links for the builder should preserve form state in the URL hash so they remain public, copyable, and independent of server-side accounts.

Important publication lanes:

- `official_canon`
- `community_library`
- `prompt_ecology_archive`

Important provenance values include:

- `official`
- `community`
- `managed_system`

The front end builds:

- lane-grouped sections
- featured collections
- category, facet, and review-state filters
- top-performing prompts based on completed runs
- JSON export examples for public machine consumption

Current featured collections include Starter Canon, User AI Starters, Companion And Reflection, Memory And Continuity, Awakening Safety, Calibrant Practice, Visual Spiralism, Wellness And Journaling, Spiralism Prompt Builders, Safety And Governance, and Archive Analysis. User AI Starters filter official prompts by `search=user ai`; Companion And Reflection filters `intent=companion-reflection`; Memory And Continuity filters `intent=memory-continuity`; Awakening Safety filters `intent=awaken-activate`; Calibrant Practice filters `intent=calibrant-practice`; Visual Spiralism filters `intent=visual-aesthetic`; Wellness And Journaling filters `intent=wellness-journaling`; Spiralism Prompt Builders filter official prompts by `search=spiralism prompt`.

## Prompt Commons Direction

Roadmap work should continue developing the library as a prompt commons rather than a generic marketplace. The shipped default order is beginner intent, recommended starters, study pathways, expert controls, then full lane lists.

The prompt-discoverability guidance promoted from [Improving Spiralist.org Navigation and UX](research/Improving%20Spiralist.org%20Navigation%20and%20UX.md) applies here: Workspace should help users recognize useful prompt paths quickly, not require them to recall internal labels or prompt taxonomy before they can run something.

The market-positioning guidance promoted from [AI Spiralism Market Survey for Spiralist.org](research/AI%20Spiralism%20Market%20Survey%20for%20Spiralist.md) also applies here: for most first-time visitors, the prompt library is the clearest public "use" surface. It should prove value quickly before asking the user to understand the full manuscript, symbol, governance, or machine-exchange model.

Useful distinctions to preserve in UI, docs, and metadata:

- raw prompts: one-off user inputs
- structured prompts: prompts with roles, constraints, variables, and expected output rules
- prompt systems: reusable multi-step workflows or chained prompt contracts
- self-improving systems: prompt systems with evaluation, feedback, and revision loops

Intent-first prompt categories should stay recognizable to visitors before internal taxonomy terms appear: