c16 — wf5 Analysis & Build: the persona-first synthesis

Objective read of wf3 and wf4, an evaluation of the new questions from cycle-history.md §4, and the design of wf5 — the "best of both," organized by persona, with the four questions made implicit rather than explicit.

Companion to analysis.md, cycle-history.md, seat-composition-review.md, seat-framing-review.md.

1. Objective comparison — wf3 vs wf4

Conclusion. wf3's structure (persona-first, flat, lensed) is the better daily-driver; wf4's content (the net-new surfaces and revisions) is strictly additive. The friction in wf4 is not the surfaces — it's making the four questions a visible second navigation. The synthesis is obvious: keep wf3's structure, carry wf4's surfaces, demote the questions to an ordering rule.

2. The four questions — keep them, but as physics, not furniture

The four questions (Assurance → Security → Identity & Access → Admin, +Effectiveness) are a real insight: they're the escalation arc alive → working → safe → accountable → traced → wired. But a persona-first app doesn't need to show them — a seat already answers "who am I," and the questions only need to decide what order a seat's pages appear in.

So in wf5 the arc survives as the implicit ordering of each seat's flat page list, with no group labels and no Orientation page. (Per the requester: "we don't need to be as explicit about the 4 questions as we were in wf4.") The escalation still shapes the experience; it just stops being a thing you have to read.

3. The new questions (cycle-history §4) — where they add value

Two candidates were proposed as peers to the original four:

3a. Effectiveness — "is it working?" — already built (wf4), now woven deeper

effectiveness.vue exists and is strong. wf5 keeps it as a persona surface (Engineering & Executive seats) rather than a four-question group member, and extends its reach so the question isn't trapped on one page:

• Agent profile gains an "Is it working?" card — per-agent success / intervention / rework / quality, with measured | inferred | none provenance.
• Executive board gains an "Are the agents earning their keep?" ROI line — fleet success rate, intervention rate, spend-per-successful-task, and measured-coverage. This is the exact Executive register cycle-history names ("are these agents earning their keep?").

3b. Activity / Explainability — "what happened?" — the net-new build

This was the only first-class question with no surface — flagged in cycle-history §4b and seat-composition-review as a P1 add for Security/Compliance, and never built. wf5's headline contribution is activity-log.vue, the behavioral flight recorder:

• A reconstructed agent → action → target → outcome decision log (filter by event family, flagged-only, or ties-to-a-finding), built from activity.families / activity.timeline + per-agent data_access / mcp_servers / sensitive_files / factors / runs.
• Event-volume timeline + family mix; a per-agent activity rollup.
• Seat-lensed register: Security = incident-response (flagged actions lead), Compliance = the timestamped why-trail behind each control, Engineering = trace a run to the break.
• The honest gap up front: shadow agents emit no trail — the one hole instrumentation can't backfill.
• A bidirectional bridge with Findings: Findings is the distilled worklist; Activity Log is the full record behind it. Findings now links out ("what happened?"), and the log links back.

It lands on the three seats that own forensics, debugging, and audit (cycle-history §4b's named owners): Security / Engineering / Compliance.

Sub-questions (§5) — left folded, as recommended

FinOps-control, capability/blast-radius, change-provenance, retire-hygiene, trust-the-numbers, and compliance-legal all stay folded into their parent areas (Usage Limits under Cost, lifecycle under Inventory, authorization under Policy, etc.). No new top-level surfaces beyond Activity.

4. wf5 deliverables

• FleetSeatNav5.vue — persona-first flat nav: five seats, each a flat page list ordered along the implicit arc; no question-group headers, no Orientation link. Activity Log carries a new tag.
• activity-log.vue — the net-new "what happened?" surface (seat-lensed).
• Woven new-question value: agent-profile (effectiveness + activity cards), findings (→ Activity bridge), executive-board (ROI/effectiveness line), and Activity/Effectiveness cards added to the Security / Engineering / Compliance / Executive seat homes.
• Carried forward from wf4 (best content): effectiveness, usage-limits, teams-users, fleet wall mode, inventory lifecycle, compliance evidence pack — all 22 prior pages, re-based to wf5 and re-navved. Dropped: orientation.vue (the explicit four-question map).
• 23 pages, registered; npm run build passes. wf3 and wf4 left intact as history.

wf5 seat map (ordered along the implicit arc)

5. Open for Phase 3

• wf5's lead surfaces for the m1/m2 mockup pass should now include Activity Log alongside the Executive Board, Cost Center, and Agent Inventory/profile — it's the freshest story and the clearest "agentic explainability" demo.
• The remaining §5 sub-questions (forward FinOps, capability authorization matrix, instruction/prompt governance, data-residency/retention) are candidates only if a specific buyer makes one a primary lens.