Agent Dashboard
Why
Qpoint is shifting as a company toward agent governance. The existing product
wireframe (wf1, in ../app.qpoint.io, running at http://localhost:7878/) was built as
an API/traffic observability platform, with only an emerging AI/agent angle (MCP clients,
dataset views, AI integration). Before investing in a redesign we need a clear-eyed read
of how well today's ~45 pages serve the stakeholders who own the agent-governance mandate
— Security, Finance, and the Executive team — and where the biggest gaps and opportunities
are, especially executive/finance overview pages and reports that don't yet exist.
Objective
Re-evaluate the wf1 app through an agent-governance lens, then explore and build a stronger version. Three phases:
- Analysis — a per-page evaluation (purpose, per-stakeholder value score 0–5, opportunities) and a per-persona evaluation (group pages by persona; assess fit, gaps, and low/high-hanging opportunities; surface missing personas and missing exec/finance overviews). Saved as a markdown report in this cycle folder.
- Wireframing — a divergent lo-fi pass exploring alternative fixes to the Phase 1 deficits (both net-new governance pages and in-place reframes); each problem gets 2–3 directions to compare, and a winning direction per problem is chosen.
- Mockups — two hi-fi passes of the winning wireframe directions: m1 built only
from existing
../q-nuxt-layercomponents, and m2 unconstrained (use/invent the best component for each job).
Scope
In
- Phase 1 analysis report grounded in the actual
../app.qpoint.iopages - Stakeholder value scoring for Security / Finance / Executive (+ reflection on other personas worth adding)
- Identification of net-new executive & finance overview/report pages
- Divergent wireframes exploring fixes to the deficits; both phases scoped from the analysis findings
- Wireframes use
Wire*components; mockups use real design-system components (Ux*,Data*,Icon*,Form*)
Out
- Rebuilding all ~45 wf1 pages (Phases 2–3 focus on the governance story)
- Changes to the real
../app.qpoint.ioapp or../q-nuxt-layerlibrary - Real data, API calls, or auth
Deliverables
phase1-analysis.md— per-page + per-persona evaluation reportinitial-plan.md— executable master plan (Phase 1 done; Phase 2 wireframing + Phase 3 mockup plan)- (Phase 2, later run) divergent
wf1wireframes underapp/pages/c15/wireframes/ - (Phase 3, later run)
m1andm2mockup passes underapp/pages/c15/mockups/
Source Material
initial-prompt.md— the originating brief../app.qpoint.io/app/pages/**— the wf1 wireframe under analysis../q-nuxt-layer— component catalog for the m1 constrained mockup passinbox/(cycle + repo root) — QControl deep-dive, PocketOS case study, artifacts
Status
- CYCLE.md created
- Phase 1 analysis report written
- initial-plan.md written (Phases 2–3)
- P1 — IA explored (4 directions: dedicated-section / dashboard-reframe / lens-overlay / persona-partitioned)
- P1 resolved → Direction D (persona-partitioned, all 5 seats, pure-D landings)
- Persona app built (
persona-app-v1) — 5 seats + 12 deep pages, every seat card live:- 6 net-new: Spend, Posture, Registry, Egress Matrix, Data-flow, Risk Report
- 6 reframes of wf1 pages: Audit Trail, Enforcement Stacks, Sensor Coverage, Agent Reliability, Dataset Views, Data Residency
- Agent-native audit (
agent-native-audit.md) — found the egress-traffic hypothesis leaked into ~10/12 wf1 pages - wf2 agent-as-actor reset scaffolded (
agent-native-v1, iteration wf2) — shared grounded modelapp/data/c15-agents.ts+ 3 spine pages: Roster, Tool & Model Use, Authorization & Guardrails (real fields, zero traffic-lens vocabulary) - wf2 completed — 7 agent-native surfaces (Roster, Tool & Model Use, Authorization, Activity, Data Exposure, Spend, Risk & Posture); each replaces/reframes a leaked wf1 page, all from
app/data/c15-agents.ts - Egress-lens scrub verified — zero traffic vocabulary (egress/bandwidth/sensor/destination/proxy) in the wf2 UI; real field names retained only as grounding citations
- Phase 2 review + page set confirmed for hi-fi
- Phase 3 m1 (constrained) mockups built
- Phase 3 m2 (unconstrained) mockups built
Visual record
2 captures — heroes, page slices, and full-page renders. Click to view.

