Skill Desk One Architecture & Diagrams

← Documentation

Skill Desk One — Architecture & Diagrams

Simple system views for business and technical audiences.

1. User Workflow

Describe Task
→
Recommend Skills
→
Safety Preflight
→
Active Workspace
→
Generate Prompt / Bundle
→
Run in User’s AI

2. Logical Architecture

Experience Layer
Task input, Skill catalog, Active Workspace, settings, downloads
Skill Intelligence
Recommendations, Skill metadata, recipes, workspace presets
Safety Layer
Preflight, secrets detection, risk scoring, confirmation/evidence rules
Target Adapter Layer
ChatGPT, Claude, Copilot, Cursor, Gemini, Codex, Universal

3. Founder OS ↔ SD1 Mission Session · v1.53

Active mission
→
One-time nonce
→
Mission Session
→
Local session health
→
Completed return
→
Verified Skill proof

v1.53 keeps the v1.43 opened / ready-to-complete / completed session state and exact nonce-matched resume path, preserves the 9 mission-specific Skill keys, and keeps the v1.52 device-local Assumption Watchlist around the mission loop. Watch items can be escalated into Custom Missions, while their assumption/evidence/review-date state is carried by the portable Founder OS backup rather than the SD1 URL. v1.53 adds review-cadence guidance around that state without changing the verification payload or storage contract. The completion payload remains the stable v1.39 contract; evidence and Company Memory are not placed in the URL.

3A. Portable Founder Memory · v1.58

Company Profile
+
Decisions
+
Mission State
+
Assumption Watchlist
+
KPI Pulse
→
Portable JSON Backup / Restore

3B. Decision + Curation Quality Gates · v1.60

Founder decision
→
Evidence + reversibility
→
Confirm / revise trigger
|
Skill candidate
→
Naming + taxonomy + metadata + safety + usability
→
Deduplicated promotion

v1.60 adds two quality gates without changing engine or persistence architecture. Founder OS reuses existing Decision Log, Company Memory, Mission Brief and Memory Intelligence state; SD1 keeps discovery All Skills separate from My Skills / SKILL.md and only promotes curated candidates after distinctness, taxonomy, metadata, safety and usability checks. The v1.39 completion-return contract and SD1 v3.9.3 remain unchanged.

3C. Search + Export Path

Search above categories
→
Live catalog filter
→
Trust status
→
Open SKILL.md / Add to My Skills / Send to AI
All / Reviewed export
→
SD1 server CSV stream
→
Excel-compatible .csv download
Reviewed Skill
→
Add to My Skills
→
sd1_my_skills_v1
→
Combined server bundle · planned
→
Upload to ChatGPT / Claude

Live JSON remains a separate catalog-source/recovery path. Individual reviewed SKILL.md files are the current portable AI-upload path. Browser-local My Skills selection may remain for convenience, but SD1 does not claim a combined Markdown download until a true server-side bundle endpoint is released.

4. Release & Rollback

Page / UI / documentation release:

Edit with WPWriter
→
Save to WordPress
→
Verify Live Page
→
Sync Docs + Release Center

SD1 engine/plugin release:

Build Versioned ZIP
→
QA + SHA-256
→
Preserve Last-Known-Good
→
Install + Activate
→
Health Check / Rollback

Current production truth: SD1 v3.9.3 is active. v3.12/v3.13 artifacts are staging only. WPVibe is backup only; WPWriter is the normal Founders/SD1 publishing path.

Release Center: /skills-releases/   •   Engine updater/status: /skills-update/