Dataface Tasks
    Heartbeat snapshot (JSON)

    Quickstart dashboards — program setup (workspace, skill, pilot process)

    IDMCP_ANALYST_AGENT-QUICKSTART_DASHBOARDS_PROGRAM_SETUP
    Statuscompleted
    Priorityp2
    Milestonem2-internal-adoption-design-partners
    Ownerdata-ai-engineer-architect
    Initiativequickstart-dashboards
    Completed bydave
    Completed2026-03-22

    Problem

    Establish a repeatable program for Dataface dashboard packs on Fivetran quickstart open-source dbt projects: workspace layout (per-repo checkouts), dft init, a connector/product research skill, documentation of default vendor dashboards and question patterns, entity/detail board linking, and a feedback loop that spawns connector-specific tasks and improves the process while executing pilots.

    Context

    • Blocked on dashboard-linking-v1 (Dashboard linking) so link matrices and entity/detail navigation are implementable without hand-built absolute URLs; landscape: Dashboard linking — research.
    • Fivetran ships quickstart dbt projects per connector; the modeled layer often mirrors source APIs and the reports/dashboards customers know from the SaaS product.
    • We want dashboard packs that feel native while strictly using what the quickstart models expose—plus explicit navigation between overview, entity, and detail boards.
    • This task is orchestration + tooling: workspace layout, a dedicated research skill (v1 at .codex/skills/quickstart-product-dashboard-research/SKILL.md), initiative spec, and two pilot tasks (Salesforce, Zendesk) that stress-test the process.
    • Spawn rule: as pilots run, file new connector tasks via just task create when the process stabilizes; do not bundle all connectors into this task.

    Possible Solutions

    • Monorepo under Dataface for all quickstart checkouts — simple pathing but messy git and huge worktree; not recommended.
    • Recommended: sibling directory (or QUICKSTART_DBT_WORKSPACE) with one clone per connector, dft init per repo, research notes under ai_notes/quickstart-dashboards/<slug>/ or linked from initiative research.
    • Optional later: scripted just quickstart-clone <url> — only after manual flow is proven in pilots.

    Plan

    1. Workspace: Pick root path; document in initiative spec; add a short README or ai_notes/.../WORKSPACE.md with clone naming and dft init commands.
    2. Skill v1: Maintain quickstart-product-dashboard-research; during pilots, revise SKILL.md when steps are wrong or missing.
    3. Pilot tasks: Unblock Salesforce and Zendesk after the workspace + skill outline is agreed; pilots may start research in parallel with workspace doc polish.
    4. Process loop: Each pilot produces research notes, board outline, link matrix, and gap list; merge learnings into skill, spec, and decisions.
    5. Spawn follow-ups: e.g. additional connector tasks once Salesforce + Zendesk retros are written.
    6. Exit: Both pilots completed or blocked with documented blockers; skill + spec updated; retro in each pilot feeding Observed process issues below.

    Observed process issues (fill during execution)

    -

    Child tasks spawned from this program

    Implementation Progress

    2026-03-22 — Program artifacts verified and ready

    Workspace layout — Documented in initiative spec: - Root: sibling directory ../quickstart-dbt-workspace/ (or $QUICKSTART_DBT_WORKSPACE) - Per-connector: git clone <url> <connector-slug> → dft init in dbt project root - Dashboard YAML lives inside each checkout, following dft init conventions - Decision recorded in decisions.md

    Research skill v1 — .codex/skills/quickstart-product-dashboard-research/SKILL.md: - 5-section checklist (native analytics, questions/roles, model mapping, board topology, M2 interlock) - Clear inputs, outputs, and exit criteria - Points to program-level methodology in research.md and templates in _templates/

    Research templates — ai_notes/quickstart-dashboards/_templates/ (7 templates): - charter.template.md, research.connector.template.md, sources.template.md, examples-index.template.md, questions.template.md, link-matrix.template.md, gaps.template.md - README documents copy-and-rename workflow into ai_notes/quickstart-dashboards/<slug>/

    Initiative docs — tasks/workstreams/mcp-analyst-agent/initiatives/quickstart-dashboards/: - index.md — initiative overview, M2 milestone, three child tasks listed - spec.md — prerequisites, scope, workspace layout, per-connector flow, quality bar, validation - research.md — 7-part methodology (charter → inventory → borrowed disciplines → examples → personas → signals → roadmap) - decisions.md — initialized; workspace layout decision added

    Pilot tasks created and ready: - Salesforce pilot — status ready, depends on this task + dashboard-linking-v1 - Zendesk pilot — status ready, depends on this task + dashboard-linking-v1

    Blocker status: dashboard-linking-v1 is in_progress with full implementation done; review feedback not yet cleared. Pilots can begin research in parallel once this program setup completes, but full board implementation awaits linking merge.

    QA Exploration

    N/A (process, workspace, and documentation; dashboard QA happens in pilot tasks via dft serve).

    • [x] QA exploration completed (or N/A for non-UI tasks)

    Review Feedback

    • [x] Review cleared