browse-tables
GitHub提供前端表格组件(GridTable)及浏览功能的开发规范,涵盖视图编辑、共享行为及大数据架构处理,明确数据流、所有权及测试流程。
Trigger Scenarios
Install
npx skills add luxury-yacht/app --skill browse-tables -g -y
SKILL.md
Frontmatter
{
"name": "browse-tables",
"description": "Change cluster or namespace views, Browse\/catalog surfaces, shared GridTable behavior, filters, pagination, large datasets, metrics columns, or refresh-backed resource tables"
}
Browse and Tables
Classify the table task before loading broad table architecture.
Task tiers
- Narrow view edit: one column, interaction, empty state, or local filter. Inspect its row producer, identity, persistence, and adjacent tests; do not inventory unrelated tables.
- Shared table behavior: GridTable, resource-table controller, adapters,
persistence, shared filtering/sorting/pagination, identity, or columns.
Inventory affected usages and read
docs/frontend/gridtable.mdplus table modes. - Architecture or large-data behavior: query ownership, global semantics,
pagination/windowing, caps, dynamic metrics, export, or select-all. Inventory
every production resource-table usage, read
docs/architecture/large-data.md, and use table modes.
Use docs/architecture/catalog.md for discovery/Browse changes,
docs/architecture/refresh-system.md for snapshot/stream contracts,
docs/frontend/live-age.md for age behavior, and
docs/architecture/resource-metrics.md for CPU/memory/utilization. Read only
the contracts selected by the change.
Ownership
- The object catalog owns discovery, existence, canonical identity, Browse
namespace metadata, and cluster listings. The
namespacesrefresh domain owns namespace LIST rows. - Typed list/table data comes from refresh snapshots and, when applicable, matching resource-stream projections.
- Frontend reads use data access or refresh orchestration. Resource tables render
through
ResourceInventoryTablewithboundedRowsSourceorbackendQuerySource; do not add a third source shape. - Shared
GridTableand column factories own common rendering behavior. - Row actions and navigation use concrete full object references.
Workflow
- Trace backend producer, scope, cap/truncation, permission behavior, cache or query owner, stream parity, and every consumer assumption.
- Classify the table mode. For tier 2/3 work, use the definitions and inventory in table modes.
- For broad work, create or update
docs/plans/<topic>.md; move durable results into the owning architecture/frontend doc when finished. - Write the failing contract test before changing global table semantics.
- Keep counts, facets, sorting, export, selection, and actions consistent with the table's actual completeness.
Focused checks
mise exec -- go test ./backend/objectcatalog ./backend/refresh/snapshot ./backend/refresh/system
mise exec -- npm run test --prefix frontend -- browse tables cluster namespace
mise exec -- npm run typecheck --prefix frontend
For broad shared-table changes, also run
mise exec -- wails3 task qc:knip and add a static/contract test preventing
unclassified production resource-table usage. Then follow the root final
validation gate.
Version History
- ee846a6 Current 2026-08-20 13:01


