Output Artifacts
Solution artifacts
The primary output is the implemented solution --- the code, configurations, tests, and documentation that the plan specified. These are written to the solution workspace.
Executor state
| File | Purpose |
|---|---|
steps.jsonl |
Per-step DoD checklist with evidence and verdicts |
ledger.md |
Human-readable ledger of per-step state and decisions |
closure.json |
Closure gate result |
Coverage report
A report showing which plan steps were completed, which spec requirements are covered, and any gaps.
AI / Machine-advantage modes
The executor supports nine machine-advantage modes:
| Mode | What it does |
|---|---|
| A.1 Wave-parallel fan-out | Parallel execution of independent steps in separate worktrees |
| A.2 Thinking-budget allocation | Scale reasoning depth with blast-radius |
| A.3 1M-context hold-all | Full plan + ledger resident in context |
| A.4 Isolated-context verification | DoD in fresh verifier (anti-self-grading) |
| A.5 Speculative dry-run | Preview next K steps in throwaway worktrees |
| A.6 Background tasks | Long actions as background Tasks |
| A.7 Cross-run memory | Distill failure signatures into Memory |
| A.8 Worktree isolation | Every parallel effector in its own worktree |
| A.9 Background sentinel | Re-assert PASSED steps after each barrier |