fix(validate): reject condition edges routed on the wrong port field

[graycyrus]

Summary

Bug B24 (engine hardening): a condition node's outgoing edges are routed exclusively on from_port (see engine::outgoing_by_port / handler_routing / collect_input) — to_port is never consulted to decide which successor fires. Nothing in validate::validate enforced that a condition node's edges actually declare their branch ("true"/"false") on from_port.

An edge authored with the branch label on to_port instead — e.g.

{from_node:"gate", from_port:"main", to_node:"yes", to_port:"true"}
{from_node:"gate", from_port:"main", to_node:"no",  to_port:"false"}

puts both edges into the same from_port group. handler_routing sees one port group with two targets and classifies it as a parallel HandlerRouting::FanOut — it drives both successors unconditionally, with no validation error and no routing-divergence warning. The condition gate silently becomes a no-op.

This surfaced live in OpenHuman: the workflow_builder agent produced exactly this malformed shape, and a flow sent an email even though its condition field resolved to false. See tinyhumansai/openhuman's companion fix (B23: builder prompt clarification) and B18 (null-arg hard gate) in the same investigation — this PR is B24, the engine-side hardening that gives B23 teeth regardless of how a graph was authored (agent, UI, or import).

Changes

• src/error.rs: new ValidationError::InvalidConditionRouting { node, from_port } variant.
• src/validate.rs: validate() now rejects any edge whose from_node is a condition node and whose from_port is not "true" or "false". A condition with only one branch wired (the common intentional gate/filter shape) is still accepted — only the value of from_port is checked, not that both are present.

Test plan

• GGML_NATIVE=OFF cargo test --lib — all 315 tests pass (24 in validate::tests, including 4 new: accepts correct from_port shape, accepts single-branch shape, rejects the exact to_port-mislabeled bug shape, rejects an arbitrary unrecognized from_port).
• Verified every existing condition-node test fixture in the crate (engine.rs, model/mod.rs, n8n_import via the OpenHuman host) already uses the correct from_port convention — no behavior change for correctly-authored graphs.

Summary by CodeRabbit

• Bug Fixes
  • Added validation for condition-node connections to ensure outgoing routes use the supported true or false branch labels.
  • Invalid routing configurations now produce a clear validation error identifying the affected node and port.
  • Valid configurations, including condition nodes with only one connected branch, continue to be accepted.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: b2ed29aa-198c-4134-852a-864dd95c5e0e

📥 Commits

Reviewing files that changed from the base of the PR and between db4165f and e8eaccb.

📒 Files selected for processing (2)

• src/error.rs
• src/validate.rs

---

📝 Walkthrough

Walkthrough

Adds structural validation for Condition node edges, requiring "true" or "false" on from_port, with a dedicated error variant and tests for valid, partial, and invalid routing shapes.

Changes

Condition routing validation

Layer / File(s)	Summary
Routing error contract src/error.rs	Adds InvalidConditionRouting with node and port details, and tests its exact display message.
Condition routing checks src/validate.rs	Rejects outgoing condition edges with invalid from_port values while accepting valid and single-branch wiring patterns.
Estimated code review effort: 2 (Simple)	~10 minutes

Possibly related PRs

• tinyhumansai/tinyflows#5: Updates condition execution routing using the same from_port semantics.

Poem

A bunny checked the branches twice,
“true” and “false” were hopping nice.
Wrong ports got turned away,
While single paths could stay.
Validation now springs bright! 🐇

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the main validation fix for condition-edge routing on the wrong port field.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

---

_{Comment @coderabbitai help to get the list of available commands.}