Session Init & Lifecycle Skill

User runs /start. Questions upfront. Everything else silent. Then seamlessly continue working.

Script: ~/.claude/skills/start/session-init.py handles all deterministic file operations. All commands output JSON. AI only handles judgment calls (questions, assessment, continuation).

Directive Capture

Capture any text after /start as DIRECTIVE. If empty, the soul purpose is the directive. The skill NEVER finishes with "ready to go" and stops. After setup, immediately begin working.

UX Contract (MANDATORY)

1. NEVER announce step names — no "Init Step 1", no "Wave 2"
2. NEVER narrate internal process — no "Detecting environment..."
3. NEVER explain what you're about to do — just ask questions, then do it silently
4. User sees ONLY: questions and seamless continuation into work
5. Batch questions — as few rounds as possible
6. No "done" message that stops — after setup, immediately begin working

Hard Invariants

1. User authority is absolute — AI NEVER closes a soul purpose. Only suggests; user decides.
2. Zero unsolicited behavior — Skill ONLY runs when user types /start.
3. Human-visible memory only — All state lives in files.
4. Idempotent — Safe to run multiple times.
5. Templates are immutable — NEVER edit files in ~/claude-session-init-templates/.
6. NEVER auto-invoke doubt agent. Only offer it.

INIT MODE

Triggered when preflight returns "mode": "init".

Step 1: Silent Pre-flight

python3 ~/.claude/skills/start/session-init.py preflight

Returns JSON with: mode, is_git, has_claude_md, root_file_count, templates_valid, template_count. If templates_valid is false, STOP with error. Set internal flags from JSON. No output to user.

Step 2: User Questions

Round 1 — Ask upfront

Use AskUserQuestion with 2 questions:

Question 1: "What is the soul purpose of this project?"

• Options: "Build a product", "Fix & improve", "Research & learn" (user can type custom)

Question 2: "How should Ralph Loop work?"

• Options: "Automatic", "Manual", "Skip"

Store as SOUL_PURPOSE and RALPH_MODE.

Round 2 — File organization (only if root_file_count > 10)

Use an Explore agent to scan root files and propose a move map. Present to user for approval. If approved, use a Bash agent to execute mkdir -p + git mv/mv operations. If skipped or root_file_count <= 10: no file moves.

Step 3: Silent Execution

Run these commands sequentially. ZERO output to user.

# 1. Bootstrap session-context + seed active context
python3 ~/.claude/skills/start/session-init.py init \
  --soul-purpose "SOUL_PURPOSE_TEXT" \
  --ralph-mode "RALPH_MODE"

# 2. Ensure CLAUDE.md has all governance sections
python3 ~/.claude/skills/start/session-init.py ensure-governance \
  --ralph-mode "RALPH_MODE"

# 3. Cache governance before /init
python3 ~/.claude/skills/start/session-init.py cache-governance

Then run /init in main thread (Claude command to refresh CLAUDE.md).

# 4. Restore governance if /init removed sections
python3 ~/.claude/skills/start/session-init.py restore-governance

Step 4: Continuation

Transition directly into work. No "session initialized" message.

• If DIRECTIVE provided: Begin working on directive immediately.
• If RALPH_MODE = "automatic": Invoke /ralph-wiggum:ralph-loop.
• Otherwise: Begin working on soul purpose. If too vague, ask ONE clarifying question, write clarified purpose back via archive --new-purpose, then begin.

RECONCILE MODE

Triggered when preflight returns "mode": "reconcile".

Step 1: Silent Assessment

Run these commands sequentially. NO output to user.

# 1. Validate session files (repair from templates if needed)
python3 ~/.claude/skills/start/session-init.py validate

# 2. Cache governance before /init
python3 ~/.claude/skills/start/session-init.py cache-governance

Run /init in main thread.

# 3. Restore governance
python3 ~/.claude/skills/start/session-init.py restore-governance

# 4. Read soul purpose + active context
python3 ~/.claude/skills/start/session-init.py read-context

Returns JSON with: soul_purpose, has_archived_purposes, active_context_summary, open_tasks, recent_progress, status_hint.

Step 2: Self-Assessment (pure reasoning, no tools)

Using the read-context JSON output, classify:

Step 3: User Interaction (conditional)

If clearly_incomplete:

No questions. Skip to continuation.

If probably_complete or uncertain:

Ask ONE question combining assessment and decision:

"Soul purpose: '[purpose text]'. [1-2 sentence assessment]. What would you like to do?"

• Options: "Continue", "Verify first", "Close", "Redefine"

If "Verify first": Dispatch doubt-agent, fold findings into re-presented question (without "Verify first" option).

If "Close" or "Redefine": Run harvest + archive:

# Check for promotable content
python3 ~/.claude/skills/start/session-init.py harvest

If harvest returns content, AI assesses what to promote (judgment call — decisions need rationale, patterns must be reused, troubleshooting must have verified solutions). Present to user for approval. After approval, append promoted content to target files via Edit tool.

# Archive old purpose, optionally set new one
python3 ~/.claude/skills/start/session-init.py archive \
  --old-purpose "OLD_PURPOSE_TEXT" \
  --new-purpose "NEW_PURPOSE_TEXT"  # omit for close-without-redefine

If "Close" without new purpose: Ask if user wants to set a new soul purpose. If declined, the archive command writes "(No active soul purpose)".

Step 4: Continuation

Transition directly into work. No "session reconciled" message.

• If DIRECTIVE provided: Begin working on directive.
• If RALPH_MODE = "automatic" (check CLAUDE.md): Invoke /ralph-wiggum:ralph-loop.
• If soul purpose just redefined: Begin working on new purpose.
• If clearly_incomplete: Pick up where last session left off using active context.
• If no active soul purpose: Ask user what to work on, write as new soul purpose via archive command.
• Otherwise: Resume work using active context as guide.

Script Reference

All commands output JSON. Run from project root.