Pages
Index Overview — Verdict Ideas

Source: qdash / (the landing page) — c68 verdict-treatment variations

Five treatments of the landing's greeting — the v2 "5 things need attention" sentence didn't land. Each variation renders verdict + queue together against the same Meridian Monday story (secrets.yaml finding · sandbox config risk · egress decision · shadow install · adapter error), because the two are one design: what the sentence says determines what the rows must repeat. React to pixels.

Variation 1

Grouped counts, not a sentence

The flat "5 things" count equates a critical finding with an adapter hiccup. Group by family instead: the verdict is a short judgment plus toned family chips (clickable filters), and the enumerating sentence disappears — the queue already says it.

Needs attention

18 of 34 agents active now · 87% of installs monitored · updated live

Security
secrets.yaml read in sofia’s Claude Code sessionhigh-severity finding · Fri 22:41 · maps to LLM06
Review finding →
Config
Sandbox disabled on marcus’s Claude Codeconfiguration finding · −20 posture points · unresolved 3 weeks
Review config risk →
Decision
Unapproved egress destination — api.telemetry-relay.io from felix’s Codexfirst seen Sun 09:17 · 3 connections since
Allow or block →
Visibility
New shadow install — Windsurf on chen-mba is unmonitoredappeared Sun 14:02 · coverage 89% → 87%
Onboard →
Monitoring
Adapter error on kai-linux — events stopped for 2 agentslast event 3h ago
Inspect →
Variation 2

The worst thing is the headline

Newspaper pattern: don't count the news, print it. The most severe item becomes the hero — concrete, immediately actionable — and everything else compresses to one "and 4 more" line that expands into the queue. Zero abstraction between login and the thing that matters most.

High-severity finding · Fri 22:41 · maps to LLM06
secrets.yaml was read in sofia’s Claude Code session
nothing suppressed — full worklist in Findings
Variation 3

The queue is the greeting

No verdict layer at all. If the queue is well-ranked, a sentence about it is double-speak — the inbox pattern: a slim status line (day, count, what's new), then the work itself. The calmest treatment, and the one that trusts the queue most.

Monday·5 need attention·3 new since Friday18 of 34 active · 87% monitored · live
Securitynew
secrets.yaml read in sofia’s Claude Code sessionhigh-severity finding · Fri 22:41 · maps to LLM06
Review finding →
Config
Sandbox disabled on marcus’s Claude Codeconfiguration finding · −20 posture points · unresolved 3 weeks
Review config risk →
Decisionnew
Unapproved egress destination — api.telemetry-relay.io from felix’s Codexfirst seen Sun 09:17 · 3 connections since
Allow or block →
Visibilitynew
New shadow install — Windsurf on chen-mba is unmonitoredappeared Sun 14:02 · coverage 89% → 87%
Onboard →
Monitoring
Adapter error on kai-linux — events stopped for 2 agentslast event 3h ago
Inspect →
Variation 4

Act · Decide · Review

Group by the verb, not the noun — the security engineer's triage loop. The verdict is three toned numbers that promise a workflow: what can't wait, what's blocked on your judgment, what should be looked at this week. Urgency lives in the grouping, so individual rows can calm down.

Act now
1
1 high-severity
Decide
1
egress destination
Review
3
1 aging 3 weeks
Act nowcan’t wait for the weekly pass
secrets.yaml read in sofia’s Claude Code sessionhigh-severity finding · Fri 22:41 · maps to LLM06
Review finding →
Decideblocked on your judgment
Unapproved egress destination — api.telemetry-relay.io from felix’s Codexfirst seen Sun 09:17 · 3 connections since
Allow or block →
Reviewthis week, before they age
Sandbox disabled on marcus’s Claude Codeconfiguration finding · −20 posture points · unresolved 3 weeks
Review config risk →
New shadow install — Windsurf on chen-mba is unmonitoredappeared Sun 14:02 · coverage 89% → 87%
Onboard →
Adapter error on kai-linux — events stopped for 2 agentslast event 3h ago
Inspect →
Variation 5

Since Friday

Time-anchored verdict: the daily driver's real question isn't "how many things?" but "what changed since I last looked?" The verdict separates new from aging — the aging clause is the conscience of the page (nothing quietly rots) — and queue rows carry NEW / AGING badges instead of re-stating severity.

Since Friday: 3 new items — and 1 aging unresolved for 3 weeks.

18 of 34 agents active now · 87% of installs monitored · updated live

Securitynew
secrets.yaml read in sofia’s Claude Code sessionhigh-severity finding · Fri 22:41 · maps to LLM06
Review finding →
Decisionnew
Unapproved egress destination — api.telemetry-relay.io from felix’s Codexfirst seen Sun 09:17 · 3 connections since
Allow or block →
Visibilitynew
New shadow install — Windsurf on chen-mba is unmonitoredappeared Sun 14:02 · coverage 89% → 87%
Onboard →
Configaging
Sandbox disabled on marcus’s Claude Codeconfiguration finding · −20 posture points · unresolved 3 weeks
Review config risk →
Monitoring
Adapter error on kai-linux — events stopped for 2 agentslast event 3h ago
Inspect →

Comparison notes

  • 1 Grouped counts — keeps a verdict layer but stops pretending five unequal things sum; chips double as filters.
  • 2 Worst-thing-first — strongest "greeting" feel; risks burying item #2 when two things are genuinely urgent.
  • 3 Queue-as-greeting — least chrome, most trust in ranking; the all-clear state needs its own design (there's no verdict line to go green).
  • 4 Act·Decide·Review — the verbs match Findings' triage workflow, so the landing and the worklist speak one language; costs vertical space.
  • 5 Since-Friday — the delta framing from ideas-round-1; "aging" gives long-ignored items a voice, which counting treatments never do.
  • All five keep the queue as a strict projection of Findings — no treatment adds landing-side triage.

Qpoint Brand Style Guide