flowchart TD plan["PLAN.md"] --> iterate["plan-iterate --ideal
refine the plan with Q&A"] iterate --> prepare subgraph prepare ["asl prepare PLAN.md"] direction TB split["session-split
session-*.md + manifest.md"] --> expand["session-expand
check every claim against the codebase"] expand --> baseline["asl baseline
record what the checks report before any change"] baseline --> preflight{"asl preflight
loop-ready?"} end preflight -- "red or yellow findings" --> fix["plan-preflight-iterate
guided Q&A repair"] fix --> preflight preflight -- "all green" --> run["asl run
next pending session, one fresh agent turn"] run --> verify{"Python runs
the Verify checks"} verify -- "fail" --> repair["repair turn
with the failing evidence"] repair --> verify verify -- "pass" --> more{"sessions
pending?"} more -- "yes" --> run more -- "no" --> done(["all-complete"]) run -. "blocked, unsafe, guard violation" .-> paused(["paused, with a named reason"]) repair -. "repair budget exhausted" .-> paused