Session Start

/session-start

Load context. Display briefing. Flag staleness. Get to work.

Steps

1. Load L0 — Session Continuity

Read NEXT.md in the project root.

2. Load L1 — Project Context (~600 tokens)

Read (skip if missing):

• Project AGENTS.md
• Vault _context.md — top 3 sections only: Summary + Key Numbers + Active Reasoning Artifacts table. STOP before Doc Index / Decision Log / Open Questions / Related sections. This is the L1 partial read rule.
• Peer Card (People/Self.md in vault or memory folder) — structured user profile with communication style, decision patterns, strengths, growth areas. Load the Communication Style and Decision Patterns tables only (~20 lines). Template provided in starter vault. Skip if missing — suggest creating one on first session.

3. Seed Activation Check

Scan [S] entries in the Active Reasoning Artifacts table:

• If activation condition appears met → surface in briefing under "Seeds Ready"
• If dormant > 90 days → flag for review
• When a seed activates, suggest: "This seed is ready. Want to /decide on it?"

4. Staleness Check

Scan [D], [I], [E] entries in the Active Reasoning Artifacts table:

First: check for expired artifacts (those with expires: YYYY-MM-DD past today's date). Flag as EXPIRED in briefing regardless of priority.

Then: priority-based staleness (for artifacts without expiry):

5. Display Briefing

## Session Briefing — {Project Name}
**Date:** {today}
**Memory:** {score}/10 | {total} artifacts | {seeds} seeds | streak: {n}

### Continue
{from NEXT.md}

### Active Decisions
{from _context.md — show critical and volatile only, up to 7}

### Seeds Ready (if any)
{[S] artifacts where activation condition appears met}

### Stale Artifacts (if any)
{list with age and invalidation trigger}

### Pending
{from NEXT.md "Decide" section}

Memory score calculation: count artifacts (0-2pts), critical decisions (0-2pts), recency (0-2pts), no stale (0-1.5pts), seeds (0-1pt), session streak (0-1.5pts). See /memento stats for full breakdown.

6. Enable Auto-Logging

Write on to ~/.memento/auto-log. This activates the Stop hook for journal entries.

To disable: user says "turn off auto-logging" → write off to the same file.

7. Do NOT Load L2/L3

Never auto-load full docs or research files at session start.

• L2 (full _context.md + referenced docs, ~2000 tok): load when L1 indicates relevance or task touches a listed doc.
• L3 (deep vault grep — Knowledge/, cross-project _context.md, MOCs, ~4000 tok): load only for cross-project research or unfamiliar domain.

8. Memento Tip (optional, once per session)

If memory score < 5, append one tip to the briefing (rotate through these):

• "We all need mirrors to remind ourselves who we are." — Run /decide to capture your next choice.
• "Remember Sammy Jankis." — Your artifacts remember so you don't have to.
• "I have to believe in a world outside my own mind." — Plant a [S] seed for ideas that aren't ready yet.
• "The world doesn't just disappear when you close your eyes." — Run /session-complete before you leave.

If score >= 5, skip the tip. The system is working — don't nag.