Align360 — Improvement Items Dashboard

Supabase Connection

Not configured

Supabase URL

Anon Key

How This Works

The Correction Loop

This dashboard is the triage center for clone quality. Items flow through a 7-step loop:

Audit → Classify → Triage Here → Push to Factory → Re-Run Blocks → Re-Test → Ship

Where Items Come From

Items are auto-generated when someone submits an expert audit. The system classifies each finding by type (voice gap, framework gap, tone drift, etc.) and maps it to the factory skill that needs to be re-run.

What To Do With Each Item

Start — Mark as in-progress. You're working on it.
Resolve — Fixed! Add notes about what changed and which version.
Skip — Won't fix. Explain why (not relevant, duplicate, deferred, etc.).
Reopen — Changed your mind? Reopen any resolved or skipped item.

Push to Factory

When items are triaged, click Push to Factory to send corrections directly into a factory run. This injects your corrections as high-priority context for the next block execution. After pushing:

Open the factory run (link provided after push)
Reset the targeted blocks
Re-execute — corrections are baked into the LLM prompt
Re-test with clone-tester to validate improvement

Factory connection requires an API key (X-Forge-Key). Expand the "Push to Factory" panel below to configure.

Keyboard Shortcuts

Ctrl + R Refresh data Esc Close modals

Known Gaps from Smoke Test (8 items)

These gaps were identified during the Pipeline Smoke Test (Mar 28). Cross-reference these when triaging audit findings. Foundation is solid for alpha — fix before beta.

Critical (3)

Hat Debate Protocol Missing

Section 13 Self-Check (Hat Debate lite) not in system prompt. Clone can't steelman-check its own advice.

clone-compiler

CTA Psychology at 10%

Offer extractor captured pricing but missed psychological triggers, urgency framing, and objection-handling patterns.

offer-extractor

Failure Recovery ~0%

Clone has no protocol for "I tried and it didn't work" scenarios. Needs recovery pathways.

clone-compiler

High Priority (5)

Steelman Incomplete

Clone doesn't push back strongly enough on weak reasoning. Needs steelman triggers from FORGE.F/.R/.G.

clone-compiler

Echo-Check Missing

No "did you actually do the thing?" follow-up mechanism. Clone takes client at face value.

clone-compiler

GOLDEN+SHARP Granularity

Clone-tester needs individual dimension scores, not just aggregate. Can't pinpoint which quality axis is weak.

clone-tester

Pattern-Breaking at 35%

Voice extractor missed humor patterns, surprise metaphors, and energy shifts. Blocked by raw transcripts.

voice-extractor

Meta-Structures at 30%

Framework extractor captured individual tools but missed cross-tool dependency chains and meta-patterns.

framework-extractor

Total

Critical

High

Medium

Low

Pending

In Progress

Resolved

Push to Factory

Configure to enable

Select items and click Load Selected to begin triage. The AI will analyze findings against GOLDEN+SHARP and FORGE+SHIFT rubrics, group by root cause, and draft correction instructions you can accept or edit.

Supabase Connection

Improvement Items

How This Works

The Correction Loop

Where Items Come From

What To Do With Each Item

Push to Factory

Keyboard Shortcuts

Known Gaps from Smoke Test (8 items)

Push to Factory

Connection Error

No improvement items found

AI Triage

Supabase Connection

Improvement Items

How This Works

The Correction Loop

Where Items Come From

What To Do With Each Item

Push to Factory

Keyboard Shortcuts

Known Gaps from Smoke Test (8 items)

Push to Factory

Connection Error

No improvement items found

AI Triage

Resolve Item

Won't Fix