Briefing

Use /briefing or natural language. Do not introduce user-facing flags for catch-up or drift. Infer the briefing posture from the request and workspace state.

Adaptive posture

Before gathering detail, read the dynamic state capsule with mcp__tars_vault__context_gaps and mcp__tars_vault__workspace_map.

Choose one posture:

Posture	Trigger
Daily	Current session cadence and user asks for today/focus
Weekly	User mentions week or planning
Re-entry	User asks to catch up, or last session was 14+ days ago
Sparse-context	Calendar/tasks exist but recent journal/transcript/intake is thin
Drift-aware	Active initiatives, overdue tasks, inbox, or index state is stale

Blend postures when needed, but keep the output short. A re-entry or sparse briefing is still just /briefing.

Data gathering

Use subagents only for independent read-heavy work. Each subagent returns bounded JSON. The main agent performs synthesis, asks any question, and saves.

Gather:

• calendar capability, if connected
• task capability, if connected
• workspace_map for active initiatives, people, decisions, tasks, inbox, and recent journal
• context_gaps for stale or missing signals
• targeted memory reads for key people and initiatives

Detailed daily, weekly, and adaptive protocols live in references/.

Output rules

• Start with the bottom line.
• Show missing integrations or missing context as uncertainty, not failure.
• End with exactly one high-leverage low-effort ask when context is stale or sparse.
• Save daily and weekly briefings to journal. Re-entry-only orientation can be shown inline unless the user provides new context.
• Do not output generic setup nags or a long command catalog.

References

• references/adaptive.md
• references/daily.md
• references/weekly.md
• references/legacy-full-protocol.md