The UI is for an AFPS power user or AI workflow operator returning to a repo-local AFPS portfolio after context loss, restart, compaction, or handoff.
UI Interview Round 2: Trust-First Source Health
The terminal assumptions manifest for trust-first-source-health was confirmed on 2026-07-11. This coverage checkpoint preserves those decisions and asks for the remaining UI emphasis choices before a concrete mockup or canonical UI packet can be designed.
Resolved Context
Product path
research/afps-tracker, resolved from the approved flow tree and UX variation artifact.
Selected branch
trust-first-source-health, approved under parent flow uf-orient-portfolio.
Primary surface
Source Health Overview: trust verdict first, affected evidence second, and route-safe path choice only after the source posture is understandable.
Prototype boundary
UI-interview may propose a static or lightly interactive mockup only. Clickable screen buildout waits for $build-ui-screens after UI approval.
AFPS Tracker
warning usableparser highsource partial
Representative source-health verdict that explains confidence and affected evidence before the path becomes routeable.
Exec Loop Tracker
blocked2 affected refs
Representative blocking posture that keeps readable portfolio context while suppressing unsafe continuation.
Assumptions Manifest
Confirm, correct, or flag each assumption. Corrections will drive the next interrogation round or the UI proposal.
The UI experiment stays inside uf-orient-portfolio: open tracker, understand the source-health verdict and affected evidence, then select a route-safe path. It hands deep provenance, diagnostics recovery, handoff/export, and boundary explanation to their owning sibling flows.
Trust-First hands safe selected-path context to verification, affected evidence to provenance inspection, blocked or contradictory state to diagnostics recovery, and inactive or promoted paths to boundary explanation.
The UI proposal should cover these states as one trust-overview experiment: loading scan, clean source health, warning/partial source health, blocked source health, affected-evidence inspection, route-safe path selection, and empty/no-AFPS state. A future experiment route can be /experiments/trust-first-source-health.
The hierarchy should lead with the repository trust verdict, confidence/freshness facts, unresolved or contradictory evidence, and affected-path count. Route-safe paths follow only after that source posture is legible.
Expected controls are inspect affected source, inspect provenance, open diagnostics recovery, select a safe path, continue to verification, and explain a boundary. Required states are empty, loading, clean, stale, partial, warning, blocked, selected, disabled, permission-denied, unresolved, contradictory, inactive, and promoted.
The visual direction should reuse the restrained dark instrument-panel language so comparison isolates hierarchy rather than styling. Status must never rely on color alone. The mockup may use fixtures and cannot commit to storage, auth, parser implementation, file watchers, analytics, deployment, command execution, export, or write-back.
Open UI Questions
What should dominate the first screen when source health is not clean?
This choice distinguishes Trust-First from both the parked Quick-Scan baseline and the later Diagnostics-First branch.
Recommended: lead with one plain-language trust verdict, then a compact evidence ledger showing confidence, freshness, unresolved references, contradictions, and affected paths. Keep diagnostics as a clearly labelled exit rather than expanding full recovery work inline.
Lead with a plain-language repository trust verdict and a compact affected-evidence ledger covering confidence, freshness, unresolved references, contradictions, and affected paths. Keep full diagnostics recovery behind a clearly labelled exit.
Agent confidence: high
How much should the trust-first layer collapse when the repository is clean?
The branch should preserve its trust thesis without making routine clean-repo orientation unnecessarily slow.
Recommended: keep a one-line verified source-health summary pinned at the top, collapse the evidence ledger by default, and promote the route-safe path list immediately below it. The user can expand the ledger without changing screens.
For a clean repository, keep a one-line verified source-health summary pinned at the top, collapse the evidence ledger by default, and show route-safe paths immediately below it with an inline expand control for evidence.
Agent confidence: high
Which layout should the first trust-first mockup use?
The layout must make trust evidence primary while keeping route-safe selection visible enough to complete the orientation job.
Recommended: desktop uses a full-width trust verdict, a two-column body with affected evidence on the left and path eligibility on the right, then a compact selected-path preview. Tablet and mobile collapse to verdict → evidence → eligible paths → selected preview.
Use a full-width trust verdict above a two-column desktop body: affected evidence on the left and path eligibility on the right, followed by a compact selected-path preview. Collapse responsively to verdict, evidence, eligible paths, then selected preview.
Agent confidence: medium
Compile Responses
Answer the assumptions and open questions, then compile YAML and paste it back to the agent.