8 criteria every automation must pass
Every automation — whether standard or lite — must pass all 8 criteria below before it counts toward bonus compensation. No partial credit.
Execute the workflow 3 consecutive times with real or realistic test data. Zero errors, zero manual intervention. If it can't run 3 times clean, it's not ready.
Every API call and external connection has an error path. If something fails mid-workflow, the automation handles it gracefully instead of silently breaking.
Builder identifies at least 2 plausible edge cases (empty data, API timeout, malformed input, rate limits) and documents how the automation handles each.
On failure, a notification fires: Slack message, email, or error log entry. Leo should never discover a broken automation by noticing missing output.
A single document (markdown or Google Doc) covering: what the automation does (1–2 sentences), trigger and schedule, inputs/outputs, error handling summary, and any dependencies or API keys required.
Builder records the n8n/Make.com operations consumed per run at acceptance time. This establishes baseline for monitoring.
Leo or designated reviewer confirms the automation does what the brief asked for. This is the human check.
If the reviewer can describe a simpler way to achieve the same result and the builder cannot justify the added complexity, the build fails this criterion.
Complexity is a risk to your bonus, not a path to more money. The incentive is to build the simplest thing that works reliably. If a 3-node workflow does the job, a 15-node workflow is not better — it's riskier.