Source: c78 fleet-overview-reports — the landing as a summary of the four reports, 2026-08-21
Tyler's organizing concept, composed: "the fleet overview, which is the landing page, is really just a summary of those 4 reports." Four cards wearing the stage-4 provisional treatment (register spine + one worst line + receipt whisper), a derived verdict line above, the estate funnel below. c68 reconciliation: the exception queue is distributed into the cards' worst lines — the verdict keeps the total; the trust band becomes the quiet footer. Every attention-state number derives from the report fixtures.
3 of your four reports need attention
posture 86 ▼4 · Mon 09:00✗CC6.1 failing — secrets.yaml read in an agent session
CMP-2026-08 · run Aug 1 open →✓F-1027 medium — open 21d against 90d target
SLA-2026-08 · run Aug 1 open →▲5 shadow installs — newest: Windsurf on chen-mba
INV-2026-08 · run Aug 1 open →✗Q3 packet will carry 1 open high unless resolved
BRD-2026-Q2 · run Jul 5 open →Composition notes
- The cascade is live: card → full report (/pages/report-*-v1) → controls / assets / specific sessions. Tyler's three levels, navigable.
- Queue reconciliation (c68): the exception queue is not a separate band here — each report card carries its own worst line, and the verdict sentence keeps the fleet-wide count. Whether c68's dedicated queue band returns above the spine is Mark's treatment pick; the mock takes the distributed position deliberately.
- Trust band reconciliation (c68): presence/coverage/estate compress to the funnel footer — the Inventory card owns coverage judgment now. The full band remains available in the c68 mocks if the compression loses too much.
- Card treatment: stage-4 provisional merge — register spine (score · delta · spark), one worst line, receipt whisper in the footer. Worst lines are shortened variants (worstShort) of the same derived strings; the full sentence lives on the report.
- Cross-report double counting (finding): one underlying event surfaces in several reports — the failing control is the open high finding, the shadow installs are the coverage gap. So the verdict counts reports needing attention, never the sum of per-report items. Any future fleet-wide item count needs entity resolution across reports first.
- Attention world derives; the clear world is authored — a second fixture world (the c68 all-clear Meridian: 39 agents, 100% coverage, nothing open), since its source reports aren't modeled. The design rule it tests: a healthy landing should read calm, not empty.
- This page is c78's deliverable to the Mark M engagement — the concrete "here's what we mean" artifact for the fleet-overview/reporting consulting scope.