ELF Checkin Command

Interactive workflow to load the building context before starting work.

What It Does

The /checkin command:

• Shows the ELF banner with ASCII art first (before any prompts)
• Queries the building for golden rules and heuristics
• Displays relevant context and frameworks
• Asks if you want to launch the dashboard (first checkin only)
• Asks which AI model you want to use (first checkin only)
• Checks for pending CEO decisions
• Loads and displays recent session context

Usage

/checkin

The checkin command is simple - just type /checkin to load framework context and prepare your session.

Execution

This skill runs the Python-based orchestrator in non-interactive mode:

python ~/.claude/emergent-learning/src/query/checkin.py --non-interactive

Handling Non-Interactive Mode

The --non-interactive flag tells the script to output [PROMPT_NEEDED] JSON hints instead of blocking on input(). When you see these in the output:

1. Dashboard prompt: [PROMPT_NEEDED] {"type": "dashboard", ...}

   • Use AskUserQuestion tool to ask: “Start ELF Dashboard?”
   • Options: “Yes (Recommended)” / “No”
   • If yes, run: bash ~/.claude/emergent-learning/dashboard-app/run-dashboard.sh

2. Model selection: [PROMPT_NEEDED] {"type": "model", ...}

   • Use AskUserQuestion tool to ask: “Which AI model for this session?”
   • Options: “Claude (Recommended)” / “Gemini” / “Codex” / “Skip”
   • Store selection for subagent routing

The orchestrator is a complete 8-step workflow:

• Step 1: Display banner
• Step 2: Load building context
• Step 3: Display golden rules & heuristics
• Step 4: Previous session summary (optional/async)
• Step 5: Dashboard prompt (first checkin only, with state tracking)
• Step 6: Model selection prompt (first checkin only, with persistence)
• Step 7: CEO decision checking
• Step 8: Ready signal

Workflow Steps (8-Step Structured Process)

Step 1: Display Banner ✓

Show ELF ASCII art immediately

• Always shown on every checkin
• Signals that framework is loading

Step 2: Load Building Context ✓

Query the learning framework

• Loads golden rules (Tier 1)
• Loads heuristics (Tier 2)
• Loads recent patterns and learnings

Step 3: Display Golden Rules & Heuristics ✓

Parse and format context for readability

• Shows rule count and key principles
• Displays relevant patterns

Step 4: Previous Session Summary

Spawn async haiku agent to summarize recent work

• Async execution (doesn’t block)
• Shows continuity with previous sessions

Step 5: Dashboard Prompt ⚡ NEW

Ask user if they want to start the dashboard

• Only on first checkin (tracked via state file)
• “Start ELF Dashboard? [Y/n]”
• Launch in background if yes
• Never asked again in same conversation

Step 6: Model Selection ⚡ NEW

Interactive prompt to select your active AI model

• Only on first checkin (state-tracked)
• Options: (c)laude / (g)emini / (o)dex / (s)kip
• Selection stored in ELF_MODEL environment variable
• Persists for subagent invocations

Step 7: CEO Decisions

Check for pending CEO decisions in ceo-inbox/

• Lists count and first 3 items
• Informational only

Step 8: Ready Signal ✓

Print completion message

• ”✅ Checkin complete. Ready to work!”
• Marks first checkin complete (state file)

Key Improvements (Full Spec Compliance)

✅ Banner First - Displayed before any prompts, not after ✅ One-Time Prompts - Dashboard and model selection appear only on first checkin ✅ State Tracking - Uses ~/.claude/.elf_checkin_state to track conversation state ✅ Model Persistence - Selection stored in ELF_MODEL environment variable ✅ Structured Workflow - All 8 steps executed in proper sequence ✅ Context Parsing - Query output properly formatted for display

Interactive Prompts

Dashboard Prompt (First Checkin Only)

Start ELF Dashboard?
   The dashboard provides metrics, model routing, and system health.

Start Dashboard? [Y/n]:

• Default: Yes (just press Enter)
• Launches in background if accepted
• Never asks again in same conversation

Model Selection Prompt (First Checkin Only)

Select Your Active Model
   Available models:
     (c)laude    - Orchestrator, backend, architecture (active)
     (g)emini    - Frontend, React, large codebases (1M context)
     (o)dex      - Graphics, debugging, precision (128K context)
     (s)kip      - Use current model

Select [c/g/o/s]:

• Stores choice in ELF_MODEL environment variable
• Used by subagent routing
• Default: Claude (s)kip option

Integration with Building

The checkin workflow is your gateway to the building’s knowledge:

• Golden Rules - Constitutional principles (always loaded)
• Heuristics - Reusable patterns and knowledge
• Failures - What went wrong and lessons learned
• Successes - What worked and can be replicated
• Sessions - Previous work summaries for continuity

Running checkin at the start of each session ensures you’re working with current institutional knowledge.