fix: workflow incorrectly marked as completed while nodes are still executing

[tomerqodo]

Benchmark PR from agentic-review-benchmarks#13

---

Note

Refines skip propagation and execution triggering in graph traversal.

• When has_taken is true, call start_execution and then enqueue_node for the downstream node
• In _propagate_skip_to_node, recurse propagate_skip_from_edge before marking each outgoing edge as skipped
• Add unit tests for SkipPropagator covering unknown/taken/all-skipped cases, recursive propagation, and branch path skipping

^{Written by Cursor Bugbot for commit c7172e3. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

[cursor]

Debug print statement left in production code

Low Severity

A debug print statement logging "Starting execution for node: {downstream_node_id}" was left in the production code. This will output to logs/stdout during workflow execution, polluting logs and potentially impacting performance in high-throughput scenarios.

 

[cursor]

Skip propagation fails due to incorrect operation order

High Severity

The order of propagate_skip_from_edge and mark_edge_skipped was reversed in _propagate_skip_to_node. When propagate_skip_from_edge runs before the edge is marked as skipped, the downstream node's incoming edge analysis sees the edge as UNKNOWN, causing has_unknown=True and premature termination. This breaks recursive skip propagation through multi-level graphs.