rworkflowsetup L2★2
WHQ25/super-one ↗What it does
Replay workflow DAG from recorded agent results
Best for
Reconstructing workflow execution DAGs from recorded results for visualization, audit, and re-execution without re-running agents; handles data-dependent fan-out and sub-workflows.
Inputs
- · [object Object]
Outputs
- · [object Object]
Requires
- · Script parser (JS/TS AST)
- · Topo order computation
- · Sub-workflow loader (via scriptPath references)
Preconditions
- · Script is valid JS (export const meta, phase(), agent() calls, parallel() calls)
- · records array contains real results from prior execution (or null if missing)
- · Data-dependent agent calls (fan-out from parallel results) are reconstructed from labels
- · Sub-workflows loaded recursively if scriptPath references present
Failure modes
- · Script syntax error → returns null (fail closed)
- · Unbounded agent loop → capped at ~5000 calls (prevents hang)
- · Recorded result missing for an agent → undefined agent result, node still recorded
- · Phase declared but spawned zero agents → placeholder phase kept (not dropped)
- · Sub-workflow reference broken (missing scriptPath) → child node recorded with subworkflow=undefined
Trust signals
- · Script is actually executable (not just inferred from AST)
- · Data-dependent fan-out reconstructed from recorded labels (not guessed)
- · Sub-workflows tagged with their name (subworkflow field)
- · Phase order preserved from script source
- · Unbounded loop protection prevents hang (call cap enforced)