Client Dashboard

Client Dashboard

When this applies

• User says "client dashboard", "status board", "show dashboard"
• User asks "who's blocked?", "who needs attention?", "what's my portfolio look like?"
• Any request for an aggregate view across all clients

Prerequisites

• clients/ directory exists with one or more {Name}-{ID}/context.md files

Workflow

Step 1 — Discover all clients

Glob for clients/*/context.md. Read every matching file in parallel (batch all reads in one tool-call round).

Step 2 — Parse each context.md

For each file, extract the fields below. Parsing must tolerate missing sections and naming variations.

Step 3 — Classify into status buckets

Assign each client to exactly one bucket, evaluated in this order (first match wins):

Step 4 — Detect flags

For each client, check for these flags. A client can have multiple.

Blocked:

• Kickoff date is > 7 days ago AND Kickoff call completed is still unchecked
• Has open action items assigned to the client contact (pattern: - [ ] {Name}: where Name is NOT "Ivan") — these are waiting on the client
• System health has critical failures for their current phase:
  • Building or later: no VAPI assistant, no phone numbers, no Responder webhook
  • Testing or later: no employee contacts, no transfer destinations

Overdue:

• A Key dates row (Testing or Go-live) has a parseable date before today AND the corresponding progress checkbox (Testing call completed or Go-live call completed) is still unchecked
• Go-live target date mentioned in the status line is before today

Stale data:

• last_synced is missing or more than 3 days before today

Action items:

• Has any open action items (the ad-hoc ones from Step 2) — note count

Step 5 — Render the dashboard

Print the dashboard directly to the user as markdown. Use this structure:

# Client Dashboard — {today's date, e.g. Apr 11, 2026}

**Portfolio:** {total} clients | {live count} live | {in-progress count} in progress | {kickoff count} in kickoff

Then render each non-empty bucket as a section, in this order: Live, Go-Live Ready, Testing, Building / In Progress, Kickoff, Paused.

For Live clients:

## Live ({count})

| Client | ID | Products | Go-live date | Status |
|--------|-----|----------|-------------|--------|
| Sykes Services | 2285 | Responder | 4/10 | Handoff pending |

Status column: note if handed off, or any open items remaining.

For Go-Live Ready clients:

## Go-Live Ready ({count})

| Client | ID | Products | Go-live target | Progress | Flags |
|--------|-----|----------|---------------|----------|-------|
| Quality Pro | 2343 | Responder | Apr 14–16 | 4/7 | 4 action items |

For Testing clients:

## Testing ({count})

| Client | ID | Products | Testing date | Progress | Flags |
|--------|-----|----------|-------------|----------|-------|

For Building / In Progress clients:

## Building ({count})

| Client | ID | Products | Next milestone | Progress | Flags |
|--------|-----|----------|---------------|----------|-------|

Progress column: {checked}/{total} from progress checkboxes. Next milestone: the first unchecked progress item. Flags: comma-separated list of flags from Step 4 (e.g. stale, blocked on client).

For Kickoff clients:

## Kickoff ({count})

| Client | ID | Products | Kickoff date | Progress | Flags |
|--------|-----|----------|-------------|----------|-------|

For Paused clients:

## Paused ({count})

| Client | ID | Reason |
|--------|-----|--------|

Skip empty buckets — don't render a section if no clients fall into it.

After the bucket tables, render an attention summary:

---

### Needs attention

List clients that have ANY flag (blocked, overdue, action items, or stale), one bullet per client:

- **{Client}-{ID}**: {comma-separated reasons}

If no clients need attention, print: All clients on track.

Then:

### Stale data (not synced in 3+ days)

- {Client}-{ID}: last synced {date}

If none are stale, omit this section.

Step 6 — Offer follow-up actions

After rendering the dashboard, offer:

1. Sync stale clients — "Want me to sync the stale clients?" (runs sync {client} for each)
2. Drill into a client — "Want details on any specific client?"
3. Prep for next call — "Want me to prep you for your next upcoming call?"

Important rules

• Read-only. Never modify any context.md file.
• Show the full board. Don't summarize or truncate — the user wants the complete picture.
• Be concise in flag descriptions. Use short phrases like stale, blocked on client, overdue: testing, 4 action items.
• Sort within buckets by urgency: clients with flags first, then alphabetically.
• Handle edge cases gracefully:
  • If a client has no System health section (e.g. pre-team clients with TBD ID), skip health-based flags for them.
  • If Key dates are blank or TBD, skip overdue detection for those milestones.
  • If Products or CRM are blank, show — in the table cell.
  • If a date format is ambiguous or unparseable, show it as-is and skip date comparisons for it.
• No auto-sync. Don't trigger the client-auto-sync rule while reading files for the dashboard. Just read and aggregate. (The auto-sync rule normally fires when reading a single client's context.md during focused work — the dashboard is a bulk read for aggregation purposes.)