Engineering evidence · Synthetic only

Six cases. Executed locally.

A small reproducible Python demonstration of the kind of failure-path reasoning behind Moruneva’s proposed reviews. We deliberately wrote a flawed mock workflow, ran six synthetic cases, and compared it with a guarded version.

Scope matters: this is a locally executed simulation with invented data. It does not test a customer workflow, n8n runtime, actual third-party API, voice agent, or production system. The intentionally flawed baseline was written for this demonstration; no real-world defects are claimed.

6 / 6Guarded simulation checks passed

0External API calls

0Customer records used

Observed results

Scenario	Deliberately flawed baseline	Guarded simulation	Check
Missing required contactmissing-input	unhandled:KeyError · 0 notifications	rejected_invalid_input · 0 notifications	PASS
Duplicate webhook delivered twiceduplicate-event	completed, completed · 2 notifications	completed, duplicate_skipped · 1 notifications	PASS
External provider times outupstream-timeout	unhandled:TimeoutError · 0 notifications	pending_retry · 0 notifications	PASS
HTTP 200 response embeds an errorfalse-ok	completed · 1 notifications	needs_review · 0 notifications	PASS
CRM write succeeds; notification failspartial-failure	completed · 0 notifications	pending_retry · 0 notifications	PASS
Normal authorized synthetic requesthappy-path	completed · 1 notifications	completed · 1 notifications	PASS

Reproduce the evidence

Run the Python source locally with Python 3 (standard library only). It simulates the six cases and writes the machine-readable results. The evidence document records the exact source SHA-256 digest so that the published output can be checked against the source.

Download runnable Python →Download recorded results

Source SHA-256: 9c25bbeddd507c17bbba3c1408268bb53e3a9ab48d83fb360e44bb59df061c99

Demonstration date: October 22, 2026. Environment: local Python 3; no third-party dependencies or external API calls.

What a real review would require

An actual buyer’s authorized, sanitized workflow or test fixture; agreed expected behavior and acceptance criteria; a permitted test environment; and a clear distinction between executed checks, static observations and tests not run. Scope and price would be agreed before work begins. No promise of finding a bug or certifying an entire production system.

Discuss an authorized pilot →Try the six-question checklist