Briefing protocol

Unified protocol for daily and weekly briefings. Mode is determined by the request signal.

Parallelization: Data gathering (calendar, tasks, memory) runs as three parallel sub-agents using the Task tool. This reduces wall-clock time for briefing generation, since calendar queries, task queries, and memory lookups are independent operations.

Daily mode

Generate a focused morning briefing for clarity and prioritization.

Steps

•
Determine date and read integrations
- •Today's date (if after 5 PM, brief for tomorrow)
- •Output folder: journal/YYYY-MM/
- •Read reference/integrations.md to determine available integrations and their status

2-4. Parallel data gathering (sub-agents)

Spawn three parallel sub-agents using the Task tool. Launch all three in a single message using multiple Task tool calls.

Sub-agent A: Fetch calendar

code

Read reference/integrations.md Calendar section for provider details.
Check calendar integration status; if not configured, return {"status": "not_configured"}.
Resolve {target_date} to YYYY-MM-DD format.
Execute the list_events operation with {target_date} and offset=1.
For each event extract: time, title, attendee names.

Return JSON:
{
  "status": "ok" | "not_configured" | "error",
  "events": [{"time": "...", "title": "...", "attendees": ["..."]}],
  "error": null | "description"
}

Sub-agent B: Fetch tasks

code

Read reference/integrations.md Tasks section for provider details.
Execute the list operation for the primary task list (default: Active).
Identify tasks due {target_date} or highest priority.
Execute the overdue operation to check for overdue tasks.

Return JSON:
{
  "status": "ok" | "not_configured" | "error",
  "tasks_due_today": [{"title": "...", "due": "...", "list": "..."}],
  "overdue": [{"title": "...", "due": "...", "list": "..."}],
  "top_priority": [{"title": "...", "due": "...", "list": "..."}],
  "error": null | "description"
}

Sub-agent C: Query memory

code

Read memory/people/_index.md.
Look up profiles for these meeting attendees: {attendee_names_if_known}
Read memory/initiatives/_index.md and find relevant initiative context.
Read reference/schedule.md if it exists; identify [RECURRING] and [ONCE] items due {target_date}.
Read inbox/pending/ and count pending items.
Read reference/.housekeeping-state.yaml for last housekeeping run.
Read reference/maturity.yaml for maturity level.

Return JSON:
{
  "people_context": [{"name": "...", "summary": "...", "open_items": ["..."]}],
  "initiative_context": [{"name": "...", "status": "..."}],
  "scheduled_items": [{"type": "recurring|once", "description": "..."}],
  "inbox_count": 0,
  "housekeeping_last_run": "YYYY-MM-DD",
  "maturity": {"level": 1, "people": 0, "meetings_processed": 0}
}

Note on attendee names: If calendar data is not yet available when spawning sub-agents (common case), spawn the memory sub-agent without attendee names. After the calendar sub-agent returns, do a quick targeted lookup for any attendees not covered. Alternatively, the memory sub-agent can load the full people index for cross-referencing after calendar results arrive.

Sub-agent input/output contracts (daily mode)

Sub-agent	Input	Output	Failure mode
Calendar	integrations.md, target date	JSON: events list with times, titles, attendees	Return `status: error`, briefing proceeds without calendar data
Tasks	integrations.md, target date	JSON: due today, overdue, top priority tasks	Return `status: error`, briefing proceeds without task data
Memory	people index, initiatives index, schedule.md, inbox/, maturity.yaml	JSON: people context, initiative context, scheduled items, system status	Return partial data, briefing uses what is available

After all three sub-agents complete, collect their JSON results and proceed to synthesis.

•
Cross-reference and enrich
- •Match calendar attendees against memory people context
- •For any attendees not covered by the memory sub-agent, do a targeted memory lookup
- •Link tasks to upcoming meetings where relevant
•
Check inbox
- •Use inbox count from memory sub-agent results
- •Offer to process pending items if any exist
•
Generate briefing

markdown

### Today's schedule
- Chronological list with time, title, prep needed

### Scheduled items due today
- Recurring and one-time items from reference/schedule.md

### Priority tasks
- Top 3-5 tasks for today
- Flag tasks that should be done before specific meetings

### People I'm meeting
- Brief context for each person
- Open items or follow-ups with them
- Questions to ask or responses owed

### Focus opportunities
- Open time slots for deep work
- Suggested task for each slot

### System status
- TARS maturity: Level [N] ([X] people, [Y] meetings). Next: [milestone]
- Inbox: [N] items pending
- Last housekeeping: [date]

•
Save and display
- •Save to journal/YYYY-MM/YYYY-MM-DD-daily-briefing.md
- •Display directly to user

Weekly mode

Generate a comprehensive weekly briefing for strategic planning.

Steps

•
Determine date range and read integrations
- •Current week: Monday through Sunday
- •Last week: Previous Monday through Sunday
- •Output folder: journal/YYYY-MM/
- •Read reference/integrations.md to determine available integrations and their status

2-4. Parallel data gathering (sub-agents)

Spawn three parallel sub-agents using the Task tool. Launch all three in a single message.

Sub-agent A: Fetch calendar (weekly)

code

Read reference/integrations.md Calendar section for provider details.
Check calendar integration status; if not configured, return {"status": "not_configured"}.
Resolve {monday_date} to YYYY-MM-DD format.
Execute the list_events operation with {monday_date} and offset=7.
Identify high-priority meetings (executives, leadership, key stakeholders).
Identify open time slots.

Return JSON:
{
  "status": "ok" | "not_configured" | "error",
  "events": [{"date": "...", "time": "...", "title": "...", "attendees": ["..."], "priority": "high|normal"}],
  "open_slots": [{"date": "...", "start": "...", "end": "...", "duration_minutes": 0}],
  "error": null | "description"
}

Sub-agent B: Fetch tasks (weekly)

code

Read reference/integrations.md Tasks section for provider details.
Execute the list operation for all configured lists (default: Active, Delegated, Backlog).
Execute the overdue operation.
Identify tasks due this week ({monday_date} through {sunday_date}).
Identify tasks related to meeting topics if known.

Return JSON:
{
  "status": "ok" | "not_configured" | "error",
  "tasks_due_this_week": [{"title": "...", "due": "...", "list": "...", "owner": "..."}],
  "overdue": [{"title": "...", "due": "...", "list": "..."}],
  "backlog_stale": [{"title": "...", "created": "...", "days_old": 0}],
  "completed_last_week": [{"title": "...", "completed": "..."}],
  "error": null | "description"
}

Sub-agent C: Query memory and context (weekly)

code

Read memory/people/_index.md + profiles of frequent meeting attendees.
Read memory/initiatives/_index.md + active initiative details.
Identify responses owed and follow-ups needed.
Read journal/{last_week_month}/_index.md for last week entries.
Read reference/schedule.md if it exists; identify items due this week.
Read inbox/pending/ and count pending items.
Read reference/.housekeeping-state.yaml for last housekeeping and index rebuild dates.
Read reference/maturity.yaml for maturity level.

Return JSON:
{
  "people_context": [{"name": "...", "summary": "...", "responses_owed": ["..."], "follow_ups": ["..."]}],
  "initiatives": [{"name": "...", "status": "...", "milestones_this_week": ["..."]}],
  "last_week_entries": [{"date": "...", "type": "...", "title": "..."}],
  "scheduled_items": [{"type": "recurring|once", "description": "...", "due": "..."}],
  "inbox_count": 0,
  "housekeeping_last_run": "YYYY-MM-DD",
  "last_index_rebuild": "YYYY-MM-DD",
  "maturity": {"level": 1, "people": 0, "meetings_processed": 0}
}

Sub-agent input/output contracts (weekly mode)

Sub-agent	Input	Output	Failure mode
Calendar	integrations.md, monday date, offset=7	JSON: events, open slots, priority flags	Return `status: error`, briefing proceeds without calendar
Tasks	integrations.md, date range	JSON: due this week, overdue, stale backlog, completed last week	Return `status: error`, briefing proceeds without tasks
Memory/Context	people index, initiatives index, journal index, schedule, inbox, system state	JSON: people context, initiatives, last week summary, system status	Return partial data, briefing uses what is available

After all three sub-agents complete, collect their JSON results and proceed to synthesis.

•
Cross-reference and enrich
- •Match calendar attendees against memory people context
- •Link tasks to meeting topics
- •Identify preparation needed for high-priority meetings
- •Match open time slots with highest priority tasks
•
Generate briefing

markdown

### Last week summary
- Achievements (completed tasks, delivered items)
- Planned but incomplete (with reasons if known)

### This week's meetings
- **High-priority**: Leadership, clients, board meetings (flag)
- **All meetings**: Chronological with date/time, title, purpose

### Open tasks
- **Due this week**: Tasks with explicit due dates
- **Meeting-related**: Tasks connected to meeting topics
- **Preparation needed**: Items requiring work before meetings

### Milestones and initiatives
- Current week milestones
- Upcoming milestones needing advance preparation
- Initiative status updates

### People context
- **Responses owed**: Promises or follow-ups to people I'm meeting
- **Questions to ask**: Follow-up items or open threads

### Recommended focus time
- Open slots matched with highest priority tasks
- Specific time blocks with suggested assignments

### Backlog review
- Backlog items older than 90 days (flag as stale)
- Suggest: keep, reprioritize, or remove

### System status
- TARS maturity: Level [N] ([X] people, [Y] meetings). Next: [milestone]
- Inbox: [N] items pending
- Last housekeeping: [date]
- Last index rebuild: [date]

•
Save and display
- •Save to journal/YYYY-MM/YYYY-MM-DD-weekly-briefing.md
- •Display key highlights to user

Progress tracking (TodoWrite)

Use the TodoWrite tool to give the user real-time visibility into briefing generation. Create the todo list at the start and update as steps complete:

Daily mode:

code

1. Determine date and read integrations                  [in_progress → completed]
2. Fetch calendar data (parallel sub-agent)              [pending → completed]
3. Fetch task data (parallel sub-agent)                  [pending → completed]
4. Query memory and context (parallel sub-agent)         [pending → completed]
5. Cross-reference and enrich data                       [pending → completed]
6. Generate and save briefing                            [pending → completed]

Weekly mode:

code

1. Determine date range and read integrations            [in_progress → completed]
2. Fetch calendar data (parallel sub-agent)              [pending → completed]
3. Fetch task data (parallel sub-agent)                  [pending → completed]
4. Query memory and context (parallel sub-agent)         [pending → completed]
5. Cross-reference and enrich data                       [pending → completed]
6. Generate and save briefing                            [pending → completed]

Parallelization note: Steps 2, 3, and 4 run concurrently as sub-agents. Mark ALL THREE as in_progress when spawning them. Mark each completed as its sub-agent returns. Do not wait for all three before updating individual statuses.

Frontmatter (both modes)

yaml

---
date: YYYY-MM-DD
title: Daily Briefing | Weekly Briefing
type: briefing-daily | briefing-weekly
---

Context budget

•Memory: Read _index.md for people and initiatives + up to 5 targeted files
•Tasks: Execute task integration list operation for Active (+ all lists for weekly mode)
•Calendar: Execute calendar integration list_events operation for today (daily) or full week (weekly)
•Journal: Current month _index.md for last week summary (weekly only)
•Inbox: Read inbox/pending/ file list
•System: Read reference/.housekeeping-state.yaml and reference/maturity.yaml

Absolute constraints

•NEVER skip calendar lookup (fall back to tasks-only if calendar integration is unreachable)
•NEVER output briefing without saving to journal
•ALWAYS resolve dates to YYYY-MM-DD format before any calendar query
•ALWAYS look up memory profiles for meeting attendees
•ALWAYS flag overdue tasks (via task integration overdue operation)
•ALWAYS check calendar integration constraints in reference/integrations.md before querying