qdash-premise-eval.md
wf3 vs. the qdash premise — fidelity & integration ideas
Evaluating app/pages/c16/wireframes/wf3/control-plane-v1/ (19 pages) against the original
purpose doc inbox/qdash — what each tab is for.md — the spec for the qcontrol/qdash app
that c16 is the agent-native re-execution of.
The doc defines 12 tabs in four question-based groups and one unifying thread:
Assurance (is the fleet healthy?) · Security (is it safe?) · Identity & Access (who's accountable?) · Admin (how is it wired up?)
One thread through all of it: honesty about what's known versus guessed.
Companion to analysis.md, phase2-review.md, seat-framing-review.md, seat-composition-review.md.
1. Verdict
wf3 is a faithful realization of the premise — and on the doc's central thesis, a more rigorous one than the doc asked for.
The doc's single unifying thread is "honesty about what's known versus guessed… the thing that makes people trust the rest." wf3 implements that thread everywhere, often more strictly than the brief:
estbadges on costs derived from token counts (cost-center), with a total-spend footnote stating what % is estimated and the line "usage-based observation, not a billed invoice."- Ghosted "not yet built" cards for the things it deliberately does not measure yet — task outcomes/quality (performance), prompt/response content (data-protection), kill switch (policy-enforcement). The gaps are shown, not hidden.
- Shadow agents priced worst-case in the posture score: "you can't claim 67 while N agents are dark" (security-posture), with a reported-vs-worst-case split.
- "Observed, not verified" on identity and showback — telemetry-inferred, not IdP-confirmed.
- "Evidence, not a conformance claim" on compliance — a finding is contributing evidence; the pass/fail is the auditor's.
So the substance is there. The one real divergence from the doc is the organizing axis, not the content (see §3.1).
2. Tab-by-tab fidelity
| # | qdash tab (group) | wf3 page | Fidelity | Note |
|---|---|---|---|---|
| 1 | Fleet Overview (Assurance) | fleet-overview | Partial | Seat-lensed, analytical, risk-first roster. Missing the doc's "calm wall screen" — ambient calm-vs-alarm at-a-glance state. |
| 2 | Agent Inventory (Assurance) | agent-inventory | Strong / Partial | Source-of-truth roster with shadow/ungoverned/compound flags. Missing the doc's lifecycle nuance: "approved / old & forgotten / nobody knew about." |
| 3 | Cost Center (Assurance) | cost-center | Strong | Spend by model/agent/user/provider, cache %, governance-tagged spend, est honesty. Matches "honest about money." |
| 4 | Performance (Assurance) | performance | Strong | p50/p95/p99, slow-AND-expensive compound target, explicit task-outcome gap callout. Matches "speedometer energy" + restraint. |
| 5 | Findings (Security) | findings | Strong | Severity worklist, ungoverned-first, framework filters. Matches "quiet is a feature; not a firehose." |
| 6 | Security Posture (Security) | security-posture | Strong | Score + fix-this-first ROI breakdown + trend + worst-case. Exactly the doc's "a score you can open up and inspect." |
| 7 | Data Protection / DLP (Security) | data-protection | Strong | Which paths / how many opens / read-write asymmetry — explicitly not contents. Exactly the doc's trust promise. |
| 8 | Policy & Enforcement (Security) | policy-enforcement | Strong | Audit→Assist→Enforce, dry-run "would have intervened on these named agents, 0 actually blocked," cost-of-flipping. Exactly "try it before you arm it." |
| 9 | Compliance & Frameworks (Identity & Access) | compliance | Strong | Behavior auto-mapped to control families, evidence-not-checkmark, evidence-gap/blind-spot framing. Export still ghosted (see idea #5). |
| 10 | Teams & Users (Identity & Access) | showback | Partial | Per-identity risk-weighted spend attribution. Missing the doc's "org chart meets dashboard" — rollups by team/owner as a first-class accountability view. |
| 11 | Agent Identity & Access (Identity & Access) | agent-identity | Strong | Credential provenance, standard conformance, rotation age, compound (weak-cred + dangerous). Exactly "treat agents as first-class identities." |
| 12 | Integrations (Admin) | integrations | Strong | Live vs projected sinks, every export warned it omits shadow agents, "X% of the fleet" coverage. Exactly "coverage as a candid headline." |
Net: 8 Strong, 3 Partial (incl. one Strong/Partial), 0 missing. Every tab exists; the gaps are refinements, not absences.
3. Integration ideas from the premise (prioritized)
Each is written to be liftable into a build plan: what it is, the premise line it serves, where it
lands, the app/data/c16-fleet.ts data that already supports it, and a rough surface shape.
3.1 — Four-question thematic layer (highest leverage, cheapest)
- What: Overlay the doc's four groups — Assurance / Security / Identity & Access / Admin — as a grouping/label layer in the nav (and optionally a one-screen orientation/sitemap).
- Premise: The doc's entire structural backbone. It organizes by question ("is it healthy/ safe/accountable/wired-up"); wf3 organizes only by persona seat ("who am I"). These are complementary — newcomers get no conceptual map from seats alone.
- Where:
app/components/wire/FleetSeatNav.vue(group the per-seat page lists under the four question-headers), or a neworientationpage. - Data: None — pure IA/labeling.
- Shape: Seat stays the primary axis; under each seat, pages cluster under their question-group label, so a user reads both "I'm in the Security seat" and "this page answers is it safe?".
3.2 — Fleet Overview "wall mode"
- What: An ambient, at-a-glance status state of
fleet-overview: calm and quiet when the fleet is healthy, visually loud when something's on fire. - Premise: "The status pane… calm when things are calm, impossible to ignore when they're not. This is the screen someone might leave up on a wall."
- Where: A mode/variant of
fleet-overview(aWireViewTogglebetween "analyst" and "wall"). - Data: Existing —
security_score/danger_leveldistribution,shadowAgents(), critical count, open-high findings. No new fields. - Shape: One dominant status verdict + the 1–3 things demanding attention; everything else recedes. The opposite of the current dense analytical layout.
3.3 — Teams & Users surface
- What: Roll agents + cost + risk + findings up by team/owner — the accountability view.
- Premise: Tab 10 — "which actual human or team is behind each agent, each cost, each risk… a problem with no owner is just a shrug… org chart meets dashboard, clean rollups by person and team."
- Where: New page in the Security/Compliance and Finance seats; distinct from
showback(which is risk-weighted spend attribution per telemetry identity, not an org rollup). - Data:
owner/user/email/org_idfor grouping; reusegovernanceTaggedSpend(),isCompoundRisk(), findings-by-agent, and the showback risk-weighting. - Shape: Group → members; per group: agents owned, spend, at-risk spend, open findings, unverified/alias count. Surfaces "this team owns 4 ungoverned agents and $3k at-risk spend."
3.4 — Agent Inventory lifecycle
- What: Add lifecycle states to
agent-inventory: active / stale / forgotten / decommission-candidate / shadow. - Premise: "which are approved, which are old and forgotten, which nobody knew about… if this list is wrong, every other tab is wrong."
- Where:
agent-inventory(a lifecycle column/filter), echoed in the roster. - Data: Existing —
last_active/first_seen(age → stale/forgotten),governed/instrumented(approved vs shadow). No new fields. - Shape: A lifecycle badge + a "stale / forgotten — last seen 90d+" filter, so cleanup candidates surface instead of hiding in the long tail.
3.5 — Compliance exportable evidence pack
- What: Promote the ghosted evidence export on
complianceto a concrete (still static) audit-artifact representation. - Premise: "Official, exportable, audit-ready. Should feel like a document you'd hand to a regulator."
- Where:
compliance(andcompliance-home). - Data: Existing findings → framework mappings, depth/breadth metrics.
- Shape: A previewable evidence-pack layout (framework → controls → contributing findings → why-trail) framed as the deliverable, not a disabled button.
4. Other things worth gleaning
4.1 — Design-feel harvest (free tone brief for Phase 3 + the Figma DS)
The doc assigns each tab an emotional design intent that the lo-fi wireframes can't carry but the
m1/m2 mockups and the new Figma design system (wzzmExswgFGXNKQyDSqGMF) should. Use this
as a per-surface tone table:
| Surface | Design feel (from the doc) |
|---|---|
| Fleet Overview | Calm when calm, impossible to ignore when not — wall-screen ambient |
| Agent Inventory | Clean, trustworthy register — "this is the source of truth," less drama |
| Cost Center | Honest about money — mark estimates as estimates or finance won't trust it |
| Performance | Speedometer energy — trends and bottlenecks easy to spot |
| Findings | A focused worklist, not a firehose — quiet is a feature |
| Security Posture | A score you can open up and inspect — breakdown as important as the number |
| Data Protection | The serious tab — weighty and clear; restraint (which/how-much, never contents) made obvious |
| Policy & Enforcement | Confidence and safety — "try it before you arm it" should feel reassuring |
| Compliance | Official, exportable, audit-ready — a document you'd hand a regulator |
| Teams & Users | Org chart meets dashboard — clean rollups by person and team |
| Agent Identity | Agents as first-class citizens with their own identities, not bolted onto the human list |
| Integrations | Coverage as a candid headline — honest rather than salesy |
4.2 — Persona-first vs question-first is a feature, not a bug
The doc is a question-first taxonomy; wf3 is persona-first. Both are valid lenses on the same fleet. Recommendation: keep seat-first and overlay the four questions (idea #1) rather than reorganize — you get "who am I" and "what question am I answering" for the cost of labels.
4.3 — The PagePurpose contract already encodes the doc's per-tab structure
security-posture ("the score, what lowers it, and where it is heading") is the build's own 5/5
template, and the WirePagePurpose shape — stake + number + direction + do-next — maps cleanly
onto the doc's per-tab what / why / who. Reaffirm it as the page contract; every new surface
above should open with it.
4.4 — Cross-check against the existing c16 reviews
- Reinforces P1 gaps already named in seat-composition-review.md: the doc's "honest about money / set budgets / catch runaway spend" backs Budgets & Spend Controls; "are they fast and working well" plus the performance task-outcome callout back Agent Reliability / Task Outcomes.
- Adds surfaces those reviews did not call out: Teams & Users (idea #3), Fleet wall mode (idea #2), and the four-question thematic layer (idea #1).
5. Recommended next step
Spin the integration ideas into a wf4 (or briefs for /wireframe): ideas #1–#4 as wireframe
work, idea #5 + the §4.1 design-feel table feeding the Phase-3 mockups (m1/m2) and the Figma
design system. This doc is the seed to expand into that plan.
Bottom line: wf3 honors the qdash premise with unusual rigor on its core "known vs guessed" thesis. The opportunities are not corrections — they are the doc's own emphases that the persona-first build under-weighted: the four-question map, the wall-screen overview, team-level accountability, inventory lifecycle, and a real audit artifact.