Goal Autopilot: the board and the engine

Understanding Goal Autopilot

When I first started using the goal-autopilot system, I wasn't sure how the goal board and engine would come together to streamline our overnight build process. The .claude/strategy/01-execution/goals/ directory contains the central board where all goal sheets are managed, while the orchestrator/ directory houses the engine scripts that automate the draining of these goals. It’s a clear division: the board is for organization, and the engine is for execution.

The Engine in Action

The real star here is the script goal_autopilot_supervisor.py, which includes commands like preflight, drain, and closeout. Each command serves a specific purpose. For example, drain is used to execute and complete goals, while closeout finalizes tasks that have met all their completion criteria. During a typical overnight run, these scripts ensure that tasks on the board are not just marked as done but are verified through a process that checks if the verifier: exits with a 0 status from the latest trunk SHA.

From Conversation to Completion

A unique aspect of the goal-autopilot system is its ability to record intake efficiently. A day of discussions can translate into a week of landed work, thanks to the structured approach to goal management. The intake process begins with reading the authority documents like FOUNDER-TRUTH-LEDGER.md and CURRENT-WEDGE.md. This ensures that every goal aligns with the broader strategic intent.

Triage and Reconciliation

Before initiating a broad drain, running goal_sheet_audit.py with --all-lanes --strict --board-hygiene provides a baseline of the current state. This step is crucial for maintaining board hygiene and ensuring that every goal is ready for execution. The goal_reconcile_landed.py script plays an essential role in identifying and resolving any discrepancies between the board and what has been completed but not yet marked.

Avoiding Common Pitfalls

Trust in the system is paramount. The DEFINITION-OF-DONE.json must be accurate, with all four rungs—BUILT, DEPLOYED, WORKS, FED—checked without skipping. The rule is clear: never manually author a DoD receipt or mark goals as done outside the prescribed transaction process. This practice ensures that the system remains reliable and transparent, avoiding the pitfalls of manual errors.

Conclusion

The goal-autopilot system, with its split ownership and precise execution routines, transforms daily conversations into actionable, verifiable outcomes. By following the structured process and respecting the trust rules, I’ve found that the board and engine work seamlessly together, turning the overnight build system into a powerful tool for progress.