Start-Day: Active Work Agent

Start-Day: Active Work Agent

Your Job

Build a unified view of all active work from memory, backlog, and the work tracking backend (DevOps Center or GitHub Issues, based on workTracking.backend). Split by assignment: current user's work, team work, and unassigned. Active work = any item that is In Progress OR (Ready + assigned).

Reference Files

• Read .claude/memory/MEMORY.md — Active Work Items table for WI status, assignment, notes
• Read docs/backlog/backlog.yaml — all items with status In Progress or Ready
• Read docs/platform-brief.md — Active Initiatives table for initiative phase context

Inputs

• Today's date: {{todayDate}}
• Current user: {{currentUserName}}
• Current SF User ID: {{currentSfUserId}}
• Quick mode: {{quickMode}}
• Production org alias: {{productionOrgAlias}}

Steps

1. Gather Active Work

Read from three sources and merge:

Source 1 — MEMORY.md: Read the Active Work Items table from .claude/memory/MEMORY.md. Extract WI number, name, backlog ID, assigned, status, and notes for each row.

Source 2 — Backlog: Read the backlog source (YAML file or GitHub Issues, depending on workTracking.backend). Filter to items where status is In Progress or Ready. For each, extract: id, title, status, assigned_to, work item references, blocked_by, priority, effort.

Source 3 — Work tracking backend (if not quickMode): Query live work item/issue status from the configured backend.

If {{quickMode}} is "true", skip the work item/issue query and report: [SKIP] Work item freshness check skipped (--quick).

If workTracking.backend == "devops-center": Query DevOps Center work items using mcp__Salesforce-DX__list_devops_center_work_items as described below.

If workTracking.backend == "github-actions": Query GitHub Issues instead:

Run: gh issue list --repo {workTracking.issueRepo} --state open --json number,title,state,labels,assignees --limit 50

Parse the JSON output:

• For each issue, extract status from labels matching status:* (e.g., status:in-progress). If no status label, use the issue state (open → "Not Started").
• Extract assignee from the assignees array.
• Match issues to backlog items by checking if the issue title or number appears in MEMORY.md or backlog context.
• Classify as "Your Active Work" (assigned to {{currentUserName}}), "Team Active Work" (assigned to others), or "Unassigned."

If the MCP query fails (server not connected, auth expired), log the failure and continue with Sources 1+2 only: [SKIP] WI freshness check — MCP unavailable: {error}

2. Merge and Classify

For each item, resolve:

• Tracking type: WI (DOC: has devops_wis entry), Issue (GHA: has Issue reference), or Backlog-only (no work item)
• Assignment: Match assigned_to against {{currentUserName}}. For DOC WI items, also cross-reference DevOps Center owner against {{currentSfUserId}}. For GHA Issues, use assignees from the Issue JSON.
• Status: Use the work tracking backend as ground truth for tracked items. Use backlog status for backlog-only items.

3. Check for Drift

Work item status drift: If backend data was retrieved, compare statuses against MEMORY.md rows (DOC) or against the backlog context (GHA). Flag any mismatches.

Assignment drift: Compare backlog assignments against the work tracking backend. Flag:

• Backlog says unassigned but work item has owner — Backlog behind
• Backlog and work item disagree on assignee — Assignment mismatch
• Backlog has assignee but work item has no owner — Work item behind

If the backend query was skipped, skip drift checks: [SKIP] Assignment drift check skipped (no work item data).

4. Read Initiative Context

Read docs/platform-brief.md Active Initiatives table. Use this to determine:

• Active initiative name and current phase
• Last session focus (from git log — check most recent commit message for context)

Output Format

Return findings in this exact structure:

### Active Work

**Active initiative:** {name from platform brief} — {phase}
**Last session focus:** {topic from most recent git commit message}

#### Your Active Work ({n} items)

| BL | Ref | Title | Status | Type | Notes |
|----|----|----|--------|------|-------|
| BL-NNNN | WI-NNNNNN | {title} | In Progress | WI | {brief context} |
| BL-NNNN | #NN | {title} | In Progress | Issue | {brief context} |
| BL-NNNN | — | {title} | In Progress | Backlog-only | {brief context} |

Use WI-NNNNNN format for DOC mode, #NN format for GHA mode. The Ref column adapts based on workTracking.idPrefix.

#### Team Active Work ({n} items)

**{Team member name}** ({n} items): {one-line summary of their focus}
- BL-NNNN: {title} ({status})

#### Unassigned Active Work ({n} items)

| BL | Ref | Title | Status | Type | Priority |
|----|----|-------|--------|------|----------|
| BL-NNNN | WI-NNNNNN | {title} | In Progress | WI | P2 |

{If unassigned In Progress items exist:}
**Note for project lead:** {n} unassigned active items — consider assigning or claiming via `/backlog update BL-NNNN`

**Pending decisions:** {list or "None"}
**Blocked items:** {list with blockers or "None"}
**Work item status drift:** {findings or "None (memory current)"}
**Assignment drift:** {findings or "None — backlog and work tracking assignments match"}

If {{currentSfUserId}} could not be resolved, skip assignment splitting: [SKIP] Assignment filtering unavailable (.env missing SF_USER_ID). Show all items in one flat table.