/Users/viktorcea/vault/Agent-System/01-Project-Rooms/Design-MD-Research/Outputs/experiments/blueprint-gl-exposure-proxy/critic-review-blueprint-base.mdHighlighted Risks
P1 items are the main design deltas the pack-guided version needs to make legible.
First-viewport review summary is passive
P1The header count says risk exists, but does not name Missing class or Blank exposure or route directly to affected rows.
Rollup review indicators are too compact
P1Issue drilldown works, but the visible rollup state is a small far-right icon instead of an inline issue label/count.
Active scope and filter state are easy to misread
P1A scoped Detail tag, a Show select, and a separate Filter popover force the user to infer which filters are active.
P1 Findings
Grouped directly from the source report.
First-viewport review summary is passive
P1The header count says risk exists, but does not name Missing class or Blank exposure or route directly to affected rows.
Rollup review indicators are too compact
P1Issue drilldown works, but the visible rollup state is a small far-right icon instead of an inline issue label/count.
Active scope and filter state are easy to misread
P1A scoped Detail tag, a Show select, and a separate Filter popover force the user to infer which filters are active.
P2 Findings
Grouped directly from the source report.
Excluded-row workflow is traceable but under-communicated
P2Exclusion is preserved and reversible, but the transition is quiet enough that users may confuse exclusion with deletion.
Detail issue semantics are diluted into long table text
P2Missing, blank, unknown, and low-confidence meanings are present, but users must scan several ordinary cells to diagnose a row.
Edit mode turns the whole table into a form
P2Full-table edit is powerful, but it does not prioritize the small set of rows that actually need correction.
Add exposure allows unresolved defaults
P2Adding unresolved data can be valid, but the baseline needs stronger validation or explicit unresolved confirmation.
Save provides no closure after correction
P2Save exits edit mode without stating whether changes were accepted or whether review items remain.
Narrow viewport pressure hides key affordances
P2At 390px, important review and correction targets can fall offscreen or become too compressed.
P3 Findings
Grouped directly from the source report.
No-results state is too generic
P3The empty state says No rows found instead of naming the active filter and next available action.
Active summary tile behavior is useful but under-signaled
P3Metric tiles navigate and expand rollups, but the selected state is subtle enough to need narration.
Acceptance Criteria
8- First load identifies active issue types and routes directly to affected Detail rows.
- Rollup review indicators show visible issue labels/counts, not icon-only warnings.
- Detail rows expose review-state tags for Missing class, Unknown state, Blank exposure, and Low confidence.
- Blank exposure displays as -- while confirmed zero displays as 0.
- Active Detail scopes, review-status filters, column filters, and result counts use one visible grammar.
- Add, save, download, exclude, restore, and delete moments provide clear local-state feedback.
- Narrow viewport behavior keeps review summary and primary correction actions reachable.
- No-results states describe the active filter and next available action.
Pack Update Recommendations
5- Require an actionable first-viewport review summary in the design-system contract.
- Add active scope plus review-status plus column-filter combinations to the component-state checklist.
- Document visible reversible excluded-row state and excluded filter fixture coverage.
- Add content patterns for Review status, Column filters, No excluded rows, save-with-review-items, and download-with-unresolved-items.
- Update the magic-trick language so review risk must become visible in the first viewport.