/nxtg-forge:ceo-loop

Level	L1 Vibe Coder
Category	CEO Decision Loop
Syntax	/nxtg-forge:ceo-loop [max-iterations] [time-limit-minutes]

---

What It Does

/nxtg-forge:ceo-loop starts an autonomous decision-making engine that runs in a continuous ORBIT cycle: OBSERVE pending decisions and project state, REASON about each decision using precedent from the decision journal, BUILD the actual decisions (approve, modify, escalate), INSPECT via retrograde analysis of previous decisions to calibrate accuracy, and TURN to prepare for the next iteration. The loop persists across Claude Code's Stop events via a hook (ceo-loop-stop.sh) that automatically triggers the next iteration.

The loop manages its own state: .claude/ceo-loop-state.json tracks iteration count, trust level, accuracy metrics, and timing. .claude/ceo-loop-decisions.jsonl is an append-only journal of every decision made. .claude/ceo-loop-progress.md bridges context across compaction events. The trust system starts at "standard" and can escalate based on decision accuracy -- more correct decisions earn higher trust, which unlocks deeper autonomous authority.

Without this command, governance decisions accumulate in .claude/ceo-decisions-pending.json waiting for manual review. The CEO loop processes them systematically, with appropriate depth for each decision type (trivial decisions are auto-approved, complex decisions get deep analysis, risky decisions are escalated to the human).

Syntax & Options

/nxtg-forge:ceo-loop [max-iterations] [time-limit-minutes]

Option	Description
max-iterations	Maximum number of ORBIT iterations to run (default: 20, range: 1-50)
time-limit-minutes	Maximum wall-clock time before the loop stops (default: 30, range: 5-120)

When to Use It

• Accumulated governance decisions: When .claude/ceo-decisions-pending.json has multiple items waiting for review.
• Project strategy session: Let the loop analyze project state, make governance decisions, and produce recommendations.
• Autonomous oversight: Start the loop and let it run through multiple iterations of observation, reasoning, and decision-making.

When you want to make a single decision manually, just review the pending decisions file directly. The CEO loop is for batch processing and autonomous governance.

Examples

Example 1: Default Loop

/nxtg-forge:ceo-loop

[CEO-LOOP] ORBIT Mode Activated
+-- Max iterations: 20 | Time limit: 30 min
+-- Pending decisions: 5
+-- Trust level: standard
+-- Starting Iteration 1 now. Stop hook will continue the loop.

The loop then runs OBSERVE-REASON-BUILD-INSPECT-TURN for each iteration. After each iteration completes, the Stop hook triggers the next one automatically.

Example 2: Custom Limits

/nxtg-forge:ceo-loop 10 60

Runs up to 10 iterations with a 60-minute time limit. Useful for longer strategy sessions.

Example 3: Resuming a Previous Session

/nxtg-forge:ceo-loop

If a previous loop was cancelled, the command detects the existing state file and decision journal. It starts a new session but the journal history informs the retrograde analysis, creating continuity across sessions.

Power Use Cases

Let the CEO loop run while you are away from the keyboard. It processes pending decisions, captures its reasoning in the decision journal, and updates the progress file. When you return, read .claude/ceo-loop-progress.md for a summary of what was decided and why.

The retrograde inspection phase is where the loop learns. It reviews previous decisions, checks if they led to good outcomes, and adjusts its trust level accordingly. Over multiple sessions, the loop builds a track record that informs future decision depth.

Combines With

Feature	Synergy
/nxtg-forge:ceo-loop-cancel	Gracefully stop the loop at any time, preserving the decision journal
/nxtg-forge:status	Check project health before starting the loop to set context
/nxtg-forge:command-center	Command center shows pending decisions that the loop can process
nxtg-ceo-loop agent	The agent that powers each ORBIT iteration within the loop

Level Progression

Level	What You Get
L1 Vibe Coder	Full ORBIT loop with decision journal, trust escalation, and retrograde analysis
L2 Pro Builder	Loop decisions integrated with orchestrator task board and knowledge base
L3 Ship Lord	Decision history and loop progress visible in the forge-ui dashboard

Tips & Gotchas

• If a loop is already active (active: true in state file), the command refuses to double-start and suggests using /nxtg-forge:ceo-loop-cancel first.
• The decision journal (.claude/ceo-loop-decisions.jsonl) is append-only and never deleted, even when the loop is cancelled. This creates a permanent audit trail.
• The Stop hook (ceo-loop-stop.sh) is what keeps the loop running between iterations. It reads the state file and re-triggers the ORBIT cycle.
• Trust levels affect decision depth: "standard" analyzes all decisions normally; higher trust levels auto-approve low-risk decisions. Trust cannot be manually set -- it is earned through accuracy.
• The progress file (ceo-loop-progress.md) is designed to survive context compaction. If Claude Code compacts context mid-loop, the hook re-feeds the progress file to restore context.

---

See also: ceo-loop-cancel | status | command-center